Chaos — When Tiny Differences Become Enormous
If mathematics tells us exactly what happens next, why can't we always predict the future?
One of the most surprising ideas students can meet beyond the normal A-level Mathematics syllabus is chaos.
At first, the word sounds distinctly unmathematical.
We normally use "chaos" to mean disorder, randomness or complete confusion. Mathematics, on the other hand, is supposed to be precise. Give a mathematician an equation and some starting values, and surely the answer should be completely predictable.
But that is not always what happens.
Some mathematical systems obey perfectly definite rules and contain no randomness whatsoever, yet their long-term behaviour can become effectively impossible to predict.
Even more remarkably, two systems starting in almost exactly the same state can eventually behave completely differently.
That is the central idea of chaos theory.
And we can investigate it with an equation simple enough to put into a spreadsheet.
A Surprisingly Simple Equation
Consider the rule:
x(n+1) = rx(n)(1 - x(n))
This is known as the logistic map.
It was originally developed from ideas about population growth, although today it is also one of the classic examples used to introduce chaotic behaviour.
There are only two important quantities.
x(n) represents the current value of the population, expressed as a fraction of some maximum possible population.
r is a parameter controlling how rapidly the population reproduces.
The equation then tells us the next value:
x(n+1).
Suppose:
x = 0.4
and:
r = 2.5
Then the next value is:
x(next) = 2.50.4(1 - 0.4)
x(next) = 2.50.40.6
x(next) = 0.6
We then feed 0.6 back into exactly the same equation to obtain the next value.
And we keep going.
This process is called iteration.
Nothing random has been introduced. Every answer is determined completely by the answer before it.
Yet something very strange is about to happen.
Experiment 1 — Build Chaos in a Spreadsheet
This makes an excellent computer-based mathematical investigation because students do not need specialised software.
Excel, Google Sheets or almost any spreadsheet will do.
Create three columns:
Iteration | x value | Second x value
In the first x column begin with:
0.5000
In the second begin with:
0.5001
The two starting conditions therefore differ by only:
0.0001
Now choose:
r = 3.9
For each new row calculate:
x(next) = 3.9x(1 - x)
Copy the formula down perhaps 50 or 100 rows.
At first, the two columns appear almost identical.
That is exactly what we would expect.
Their starting values were almost identical.
But keep going.
Something remarkable happens.
Around iteration 20, the difference is becoming noticeable.
By about iteration 25, the two values can already differ by several hundredths.
By iteration 30, one calculation can give approximately:
0.973
while the other gives approximately:
0.284
We started with:
0.5000
and:
0.5001
A difference of just one ten-thousandth.
Thirty iterations later the two systems can be in completely different places.
Nothing random was added.
Both calculations followed precisely the same mathematical rule.
Only their initial conditions were slightly different.
That is one of the defining characteristics of chaos:
sensitive dependence on initial conditions.
Draw the Two Curves
The effect becomes even clearer if the spreadsheet results are plotted.
Put iteration number on the horizontal axis and x on the vertical axis.
Plot both calculations on the same graph.
For the first few iterations, the lines may appear to lie almost exactly on top of one another.
Then they begin to separate.
Soon afterwards they appear to have almost no relationship at all.
This is far more powerful than simply telling students that chaotic systems are sensitive to starting conditions.
They can actually watch predictability disappear.
But the Equation Hasn't Changed
This raises a fascinating question.
Why should prediction become difficult?
At iteration 30, we are still doing exactly the same calculation:
x(n+1) = 3.9x(n)(1 - x(n))
There is no dice throw.
There is no random-number generator.
There is no hidden choice being made by the computer.
If we know x(n) exactly, we can calculate x(n+1) exactly.
The system is therefore deterministic.
Yet long-term prediction becomes extraordinarily difficult.
This leads to one of the most important distinctions students can meet:
Deterministic does not necessarily mean predictable.
Those two words are not synonyms.
Now Change r
There is another wonderful feature of the logistic map.
Instead of changing the starting value, keep the starting value the same and slowly change r.
The character of the entire system changes.
For relatively small values of r, the population settles down.
Increase r and oscillation begins.
Increase it again and the oscillation becomes more complicated.
Eventually the behaviour becomes chaotic.
A rough journey looks like this.
At r = 2.5 — Stability
Begin with almost any sensible starting value between 0 and 1.
After several iterations, the values settle towards a fixed value.
The population reaches an equilibrium.
The next generation is approximately the same size as the previous generation.
Everything appears reassuringly predictable.
At r = 3.2 — Oscillation
Increase r and the fixed equilibrium loses its stability.
Instead of settling at one value, the system begins to alternate between two values.
High.
Low.
High.
Low.
The population has entered a period-2 cycle.
Increase r Again — Period Doubling
Increase r further and something even stranger happens.
Instead of cycling through two values, the system begins cycling through four.
Then eight.
Then sixteen.
The cycles double again and again.
This is called period doubling.
The intervals between these changes become progressively smaller.
Eventually, at around:
r = 3.57
the behaviour becomes predominantly chaotic.
A very simple nonlinear equation has travelled from stability to oscillation to apparently irregular behaviour.
The Bifurcation Diagram
One of the most beautiful pictures in modern mathematics emerges if we repeat this experiment for thousands of different values of r.
For each r value, we discard the early iterations and plot the values that remain.
The resulting picture is called a bifurcation diagram.
At first there is a single branch.
Then it divides into two.
Those two divide into four.
Four divide into eight.
The branches become increasingly dense until a complicated region of chaotic behaviour appears.
But even inside the chaos there are unexpected windows of order.
Stable cycles suddenly reappear.
Then they split again and return to chaos.
The result looks almost like a mathematical tree growing out of a single line.
It is a spectacular reminder that very complicated structures can emerge from extremely simple rules.
Where Does the Complexity Come From?
The logistic equation contains a crucial feature:
x*(1 - x)
This makes the equation nonlinear.
Nonlinear systems behave differently from the straight-line relationships students meet early in mathematics.
For example:
y = 3x + 2
is linear.
Double x and the effect on y is straightforward.
But in the logistic map, x is multiplied by another expression containing x.
The system also feeds its own output back into itself.
Today's result becomes tomorrow's input.
A small difference can therefore alter the next result.
That altered result then creates another difference.
That difference affects the following calculation.
And the process continues.
In some chaotic systems, uncertainty effectively grows with each iteration.
Eventually the uncertainty can dominate the prediction.
The Butterfly Effect
Chaos theory is frequently associated with the phrase:
the butterfly effect.
It is sometimes exaggerated into the claim that a butterfly flapping its wings directly causes a hurricane.
That is not really the point.
The important idea is that in a sufficiently sensitive system, an extremely small difference in the initial conditions can eventually contribute to a very large difference in the final state.
Imagine trying to measure the atmosphere.
We might measure:
temperature;
pressure;
humidity;
wind speed;
wind direction.
But we can never measure every quantity at every location with infinite precision.
Suppose the true temperature somewhere is:
17.263847... degrees C
but our instrument records:
17.26 degrees C.
For many calculations that difference is irrelevant.
In a chaotic dynamical system, however, tiny differences can grow.
This places a fundamental limit on how far ahead some systems can realistically be predicted.
Why Weather Forecasts Become More Difficult
Weather is an excellent real-world connection, although the real atmosphere is vastly more complicated than the logistic map.
Modern forecasting begins with observations of the atmosphere and uses mathematical models to calculate how conditions are likely to evolve.
But we never know the exact state of the entire atmosphere.
There are always uncertainties.
Because atmospheric dynamics can display chaotic behaviour, forecasts that begin with very slightly different initial conditions can eventually diverge.
This is one reason meteorologists use ensemble forecasts.
Instead of running only one simulation, computers run many forecasts beginning from slightly different plausible starting conditions.
If nearly all the simulations produce a similar outcome, confidence may be relatively high.
If the simulations spread widely, uncertainty is greater.
Chaos therefore does not mean:
"weather forecasting is impossible."
It means there are limits to how precisely some aspects of weather can be predicted far into the future.
Populations — Where the Logistic Map Began
The logistic map is particularly interesting because it can be interpreted as a simplified population model.
Imagine a species in an environment with limited resources.
If the population is small, there is plenty of food and space.
The population can grow rapidly.
But as the population increases, competition becomes stronger.
Growth is restricted.
This is represented by the factor:
(1 - x)
When x is small, this factor is large.
When x approaches 1, it becomes small.
The equation therefore contains both reproduction and limitation.
Of course, real ecosystems involve predators, disease, migration, climate, age structure and countless other factors.
The logistic map is not a realistic complete model of an ecosystem.
Its importance is that even a drastically simplified deterministic population model can produce extraordinarily complicated behaviour.
A Double Pendulum — Chaos You Can See
There is also a wonderful physical demonstration of chaos.
Take an ordinary pendulum and attach a second pendulum to its end.
The result is a double pendulum.
Release it from one position and record the motion.
Then reset it as accurately as possible and release it again from almost the same position.
Initially the two motions may look similar.
Soon they can become completely different.
The movement can become spectacular: swinging, rotating and reversing direction in ways that are extremely difficult to anticipate.
Again, the motion is governed by physical laws.
The pendulum is not deciding randomly where to move.
But the system is highly sensitive to its starting conditions.
It would make an excellent companion practical to the spreadsheet experiment.
One is mathematical.
One is physical.
Both demonstrate the same underlying idea.
Turbulence and Fluid Motion
Another connection appears in moving fluids.
Water flowing slowly through a pipe can display relatively orderly behaviour.
Increase the speed and the motion may become turbulent.
Eddies form within eddies.
Structures appear, change and disappear.
Air flowing around buildings, aircraft wings, sails and vehicles can show similarly complicated behaviour.
Turbulence is a far more complicated subject than the logistic map, and the two should not simply be treated as the same thing.
But both belong to the wider mathematical world of nonlinear dynamical systems, where simple expectations about cause and effect can fail.
Does Chaos Mean Everything Is Random?
No.
This is perhaps the most important misconception to challenge.
Random behaviour involves genuine unpredictability in the process or a probabilistic description of outcomes.
A chaotic deterministic system follows fixed rules.
If we somehow knew the starting conditions with infinite accuracy and could perform the calculations with infinite precision, the future state would be determined.
The problem is that real measurements do not contain infinite information.
Neither do computers.
Numbers have to be stored to finite precision.
And if tiny errors grow rapidly, eventually those tiny uncertainties matter.
Chaos therefore creates a fascinating middle ground.
The system is not random.
But its long-term behaviour can become practically unpredictable.
A Very Good Student Challenge
Once students have created the spreadsheet, I would encourage them not simply to accept the standard values for r.
Explore.
Try:
r = 2.0
r = 2.8
r = 3.1
r = 3.4
r = 3.5
r = 3.55
r = 3.6
r = 3.8
r = 3.9
r = 4.0
For each value, ask:
Does the system approach a single value?
Does it oscillate?
How many values appear in the cycle?
Does it appear chaotic?
How much does changing the initial value matter?
How many iterations are required before two nearby starting conditions noticeably separate?
Then try changing r in much smaller steps.
Students will begin discovering the bifurcations for themselves.
That turns the exercise from a demonstration into a genuine mathematical investigation.
Go Further — Can You Find Order Inside Chaos?
There is an additional surprise for students who want to explore further.
Chaotic behaviour does not simply begin and then continue uniformly.
Within chaotic regions there are windows of periodic behaviour.
For certain values of r, an apparently chaotic system suddenly settles into a repeating cycle again.
Then that cycle undergoes its own sequence of period doubling.
So even inside apparent disorder, mathematical structure remains.
That is one of the reasons the bifurcation diagram is so fascinating.
It is not simply a picture of increasing messiness.
It contains extraordinary organisation.
An Even Deeper Result — The Feigenbaum Constant
There is another beautiful piece of mathematics hiding here.
As the period doublings occur, the spacing between successive bifurcations decreases in a systematic way.
The ratio between these intervals approaches approximately:
4.669...
This number is known as the Feigenbaum constant.
The remarkable thing is that the same constant appears in many completely different nonlinear systems undergoing period doubling.
So a pattern first explored through a simple population equation reveals something much deeper.
Different mathematical and physical systems can approach chaos in remarkably similar ways.
That idea — that apparently unrelated systems can share universal mathematical behaviour — is one of the great attractions of mathematics beyond the examination syllabus.
Why I Like This as an A-Level Investigation
There is very little difficult calculation here.
An A-level student can understand the equation.
A spreadsheet can perform the repeated arithmetic.
Yet the ideas lead rapidly into university-level mathematics, physics, meteorology and computational modelling.
That makes it exactly the sort of topic I enjoy exploring beyond the syllabus.
Students sometimes assume that more advanced mathematics must mean longer equations, more complicated algebra and increasingly difficult manipulation.
Chaos theory demonstrates something much more interesting.
A simple equation does not necessarily produce simple behaviour.
In fact, one of the deepest questions becomes:
How much can we know about the future even when we know the rules?
Mathematics, Measurement and Prediction
There is also a useful scientific lesson here.
Whenever we make a prediction, three things matter:
the mathematical model;
the starting information;
the sensitivity of the system to errors in that information.
If a system is not particularly sensitive, small measurement errors may remain small.
Prediction can remain useful for a long time.
If the system is chaotic, an apparently insignificant uncertainty can eventually grow until two possible futures bear little resemblance to one another.
That distinction matters in fields ranging from weather forecasting to orbital dynamics, fluid mechanics and biological populations.
The Bigger Lesson
Students are often introduced to mathematics as a subject in which every problem has a definite answer.
And in one sense, the logistic map reinforces that idea.
At every stage we can calculate exactly what the next number should be.
But it also reveals something deeper.
Knowing the rule is not always enough to make useful long-term predictions.
A system can be:
deterministic but unpredictable;
simple in its rule but complicated in its behaviour;
orderly in one region and chaotic in another;
extraordinarily sensitive to differences too small to notice initially.
That is a much richer view of mathematics.
Conclusion — Can We Predict the Future?
Start with:
x = 0.5000
and:
x = 0.5001.
The difference appears insignificant.
Apply exactly the same deterministic mathematical rule repeatedly.
At first the answers stay close.
Then they separate.
Eventually they can describe entirely different states of the system.
No randomness was introduced.
No rules were changed.
Nothing went wrong with the mathematics.
The unpredictability emerged from the mathematics itself.
That is the extraordinary lesson of chaos theory.
We often imagine that if we know the laws governing a system, we should be able to predict its future.
Chaos tells us something more subtle:
Knowing what happens next does not necessarily mean we can know what happens much later.
And all of that can begin with one surprisingly simple equation:
x(n+1) = rx(n)(1 - x(n))
