Introduction

Up to this point, measure theory has mostly studied static objects:

• Measures
• Functions
• Integrals
• Dimensions
• Fractals

Now we introduce motion.

Suppose we repeatedly apply a transformation:

$$T:X\to X$$

What happens after:

• 10 iterations?
• 100 iterations?
• 1 million iterations?

Can long-term behavior be predicted?

Can randomness emerge from deterministic rules?

These questions lead to Ergodic Theory.

Ergodic theory is one of the most important bridges connecting:

• Measure Theory
• Probability
• Dynamical Systems
• Statistical Mechanics
• Operator Algebras

and eventually:

• Alain Connes
• von Neumann algebras
• Noncommutative Geometry

---

The Fundamental Setup

Start with a measure space:

$$\left(X,\mathcal F,\mu\right)$$

and a transformation:

$$T:X\to X$$

Think of:

$$T$$

as one step of evolution.

Examples:

• One day passing
• One iteration of a map
• One update of a system

Repeated application gives:

$$x,T(x),T^2(x),T^3(x),\ldots$$

This sequence is called the orbit of:

$$x$$

---

Example 1: Rotation of a Circle

Let:

$$X=[0,1)$$

and define:

$$T(x)=x+\alpha \pmod 1$$

Each iteration rotates the circle.

If:

$$\alpha=\frac14$$

the orbit repeats.

If:

$$\alpha=\sqrt2-1$$

the orbit never repeats.

Completely different behaviors emerge.

---

Example 2: Doubling Map

Define:

$$T(x)=2x \pmod 1$$

Examples:

$$0.2\to0.4\to0.8\to0.6\to0.2$$

This map appears simple.

Yet it exhibits surprisingly complex behavior.

It is one of the standard examples in ergodic theory.

---

Measure Preservation

Not every transformation is interesting.

The crucial condition is:

Measure Preservation

A transformation:

$$T$$

is measure-preserving if:

$$\mu(T^{-1}(A))=\mu(A)$$

for every measurable set:

$$A$$

---

Interpretation

Measure preservation means:

The transformation reshuffles mass without creating or destroying it.

Think of:

• Shuffling cards
• Stirring coffee
• Mixing paint

The total amount remains unchanged.

---

Example

The doubling map:

$$T(x)=2x \pmod 1$$

preserves Lebesgue measure:

$$\lambda$$

Thus:

$$\lambda(T^{-1}(A))=\lambda(A)$$

for every measurable set.

---

Dynamical Systems

The triple:

$$\left(X,\mathcal F,\mu,T\right)$$

is called a measure-preserving dynamical system.

This is the fundamental object of ergodic theory.

Everything that follows studies the behavior of such systems.

---

Time Averages

Suppose:

$$f:X\to\mathbb R$$

is measurable.

Observe a point:

$$x$$

as it evolves.

We record:

$$f(x),f(Tx),f(T^2x),\ldots$$

The average over time is:

$$\frac1n\sum_{k=0}^{n-1}f(T^k x)$$

This is called the time average.

---

Space Averages

Independently, we can compute:

$$\int_X f,d\mu$$

This is the space average.

---

The Central Question

Are these quantities related?

Does:

$$\frac1n\sum_{k=0}^{n-1}f(T^k x)$$

eventually approach:

$$\int_X f,d\mu$$

?

This is the central problem of ergodic theory.

---

Motivation from Physics

Imagine:

$$f(x)=$$

temperature.

The time average measures:

Average temperature observed by one particle.

The space average measures:

Average temperature throughout the whole system.

Physicists desperately wanted these quantities to be equal.

This is where ergodic theory began.

---

The Ergodic Hypothesis

In the 1800s, physicists proposed:

Long-term observation of one particle should reveal the same statistics as observing the entire system.

This idea became known as the Ergodic Hypothesis.

Measure theory eventually made it rigorous.

---

Invariant Sets

A set:

$$A$$

is invariant if:

$$T^{-1}(A)=A$$

Once a trajectory enters:

$$A$$

it never leaves.

Invariant sets are fundamental.

---

Example

For the identity map:

$$T(x)=x$$

every set is invariant.

This system has almost no dynamics.

---

Why Invariant Sets Matter

Invariant sets reveal hidden structure.

If many invariant sets exist,

the system decomposes into independent pieces.

If few exist,

the system behaves more uniformly.

---

Definition of Ergodicity

A measure-preserving system is ergodic if every invariant set satisfies:

$$\mu(A)=0$$

or

$$\mu(A)=1$$

Only trivial invariant sets exist.

---

Intuition

Ergodicity means:

The system cannot be decomposed into smaller independent subsystems.

Everything eventually mixes together.

---

Example: Non-Ergodic System

Consider:

$$T(x)=x$$

Every subset is invariant.

The system is not ergodic.

---

Example: Circle Rotation

If:

$$\alpha$$

is irrational,

then:

$$T(x)=x+\alpha \pmod1$$

is ergodic.

The orbit eventually spreads throughout the circle.

---

The Birkhoff Ergodic Theorem

This is one of the greatest theorems in mathematics.

Let:

$$\left(X,\mathcal F,\mu,T\right)$$

be measure-preserving.

Let:

$$f\in L^1(X)$$

Then:

$$\lim_{n\to\infty}\frac1n\sum_{k=0}^{n-1}f(T^k x)$$

exists for almost every:

$$x$$

---

Even More Remarkably

If the system is ergodic, then:

$$\lim_{n\to\infty}\frac1n\sum_{k=0}^{n-1}f(T^k x)=\int_Xf,d\mu$$

for almost every:

$$x$$

---

Why This Is Amazing

The theorem says:

Time averages equal space averages.

A single trajectory reveals global information.

This is exactly what physicists hoped would be true.

---

Connection to the Law of Large Numbers

Compare:

$$\frac1n\sum_{k=1}^{n}X_k$$

in probability theory.

The Law of Large Numbers says this converges to:

$$E[X]$$

Birkhoff’s Theorem is a vast generalization.

Many probabilistic laws are special cases of ergodic theory.

---

Mixing

Ergodicity is only the beginning.

A stronger property is mixing.

A system is mixing if:

$$\mu(T^{-n}(A)\cap B)\to\mu(A)\mu(B)$$

as:

$$n\to\infty$$

---

Interpretation

Initially:

$$A$$

and:

$$B$$

may be strongly related.

After many iterations:

the relationship disappears.

The system behaves randomly.

---

Why Mixing Matters

Mixing explains:

• Thermal equilibrium
• Random-looking behavior
• Chaos

Many deterministic systems become statistically indistinguishable from random systems.

---

Ergodic Theory and Probability

Ergodic theory often studies deterministic systems that behave randomly.

This is one of the great surprises of mathematics.

Simple equations can generate:

• Randomness
• Chaos
• Statistical laws

without any actual randomness being introduced.

---

Connection to von Neumann

Ergodic theory became deeply connected to operator theory through the work of:

John von Neumann

He realized that dynamical systems can be represented by operators acting on Hilbert spaces.

This observation helped create:

• Operator algebras
• von Neumann algebras

---

Why This Matters for Connes

This lesson marks the first major step toward Alain Connes.

The progression is:

Measure Theory

$$\longrightarrow$$

Ergodic Theory

$$\longrightarrow$$

Operator Theory

$$\longrightarrow$$

von Neumann Algebras

$$\longrightarrow$$

Noncommutative Geometry

Connes’ early work was deeply influenced by ergodic theory.

Many von Neumann algebras arise directly from measure-preserving dynamical systems.

---

Key Concepts Learned

By the end of this lesson you should understand:

• A measure-preserving system is:

$$\left(X,\mathcal F,\mu,T\right)$$

• Measure preservation means:

$$\mu(T^{-1}(A))=\mu(A)$$

• Orbits are:

$$x,T(x),T^2(x),\ldots$$

• Ergodicity means invariant sets have measure:

$$0$$

or:

$$1$$

• The Birkhoff Ergodic Theorem states that time averages converge.
• In ergodic systems, time averages equal space averages.
• Mixing is a stronger form of randomness.
• Ergodic theory forms one of the major roads toward operator algebras and Connes’ work.

---

Looking Ahead

Measure Theory Lesson 37: The Koopman Operator

The next lesson is where measure theory starts transforming into functional analysis.

Instead of studying the transformation:

$$T$$

directly, we will study how it acts on functions.

This leads to the Koopman Operator, one of the key ideas connecting dynamical systems, ergodic theory, Hilbert spaces, spectral theory, and ultimately von Neumann algebras. From this point onward, the road toward Alain Connes becomes much more visible.