Introduction to Ergodic Theory - Core Definitions

Ergodic theory studies the statistical properties of dynamical systems, providing a bridge between deterministic dynamics and statistical mechanics. It addresses fundamental questions about long-term averages, invariant measures, and the relationship between time evolution and ensemble distributions.

DefinitionMeasure-Preserving Transformation

Let $(X, \mathcal{B}, \mu)$ be a probability space. A measurable map $T: X \to X$ is measure-preserving if for every measurable set $A \in \mathcal{B}$ :

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

Equivalently, for any integrable function $f$ :

$\int_X f(T(x)) \, d\mu(x) = \int_X f(x) \, d\mu(x)$

The measure $\mu$ is invariant under $T$ . Measure-preserving systems model conservative dynamics where total probability is conserved over time.

Measure preservation is the probabilistic analog of volume preservation in Hamiltonian mechanics. It ensures that distributions don't concentrate or disperse over time but maintain their statistical properties. This property is fundamental for applying ergodic theory to physical systems.

DefinitionErgodicity

A measure-preserving transformation $(X, \mu, T)$ is ergodic if the only $T$ -invariant sets have measure 0 or 1. Equivalently, $T$ is ergodic if:

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

for almost every $x$ and all integrable $f$ . Time averages equal space averages, making the system indecomposable into invariant subsystems.

Ergodicity means the system explores its entire phase space uniformly over time. No proper subset can trap trajectories indefinitely. This property justifies statistical mechanics' ensemble interpretation: a single long-time trajectory provides the same statistics as many initial conditions evolved briefly.

DefinitionMixing

A measure-preserving system is (strongly) mixing if for any measurable sets $A, B$ :

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

This means correlations decay to zero: initially separated distributions become independent asymptotically. Mixing is stronger than ergodicity—mixing systems are ergodic, but not conversely.

Weak mixing is a intermediate property: some subsequence $n_k$ satisfies the mixing condition.

ExampleRotation on the Circle

Consider $T_\alpha: S^1 \to S^1$ , $T_\alpha(\theta) = \theta + \alpha \pmod{2\pi}$ with Lebesgue measure:

If $\alpha/(2\pi)$ is irrational: $T_\alpha$ is ergodic (by Weyl's equidistribution theorem)
If $\alpha/(2\pi)$ is rational: not ergodic (orbits are periodic, confined to finite sets)
Never mixing (isometries cannot decorrelate)

This demonstrates that ergodicity depends crucially on arithmetic properties of parameters.

Remark

The hierarchy of properties is: $\text{Mixing} \Rightarrow \text{Weak Mixing} \Rightarrow \text{Ergodic} \Rightarrow \text{Measure-Preserving}$

Each implication is strict: there exist ergodic non-mixing systems (irrational rotations), weakly mixing non-mixing systems, and measure-preserving non-ergodic systems (rational rotations). This hierarchy organizes systems by increasing degrees of randomness and statistical regularity.

Ergodic theory provides the mathematical foundation for statistical mechanics and thermodynamics. It explains how macroscopic laws emerge from microscopic dynamics, why irreversibility appears despite time-reversible equations, and how deterministic systems can exhibit random-like statistical behavior.