Ergodicity and Time Averages: When One Trajectory Represents the Whole

Ergodicity and Time Averages: When One Trajectory Represents the Whole

How to use this page inside the site

If you want the project’s formal spine and checkable statements, use Rigidity & Reconstruction. For the structured reading map and verification paths, use Research Library.

This writing section exists to make technical words usable. Cross-domain parallels are provided as intuition, not as proof. The boundary rule is stated here: Illustrations, Not Proof.

This page explains what ergodicity is trying to guarantee: that time averages reflect the long-run statistical picture.

Ergodicity is a word that often gets used as a badge of seriousness. The useful meaning is simple: if a system is ergodic, then following one typical trajectory for a long time gives you the same averages you would get by sampling the whole state space according to the system’s invariant distribution.

Time averages vs ensemble averages

Time average means “average along one run.” Ensemble average means “average across the distribution of states.” Ergodicity is the condition that makes these two agree for typical starting points.

Why this matters

When ergodicity holds and mixing is strong, you can justify replacing long-time observation by statistical expectation. When it fails, long-time behavior can depend strongly on initial conditions or on hidden invariant components.

Connections to mixing and spectral gap

Mixing is a stronger “forgetting” property than ergodicity. Spectral gap is one way to quantify fast decay of correlations. If you want those terms in plain language, use Mixing and Relaxation and Spectral Gap.

Where to go next

If you want the broader “settling down” vocabulary, read Convergence vs Equilibrium. If you want the information/entropy vocabulary that often accompanies statistical thinking, read Entropy, Information, and the Arrow of Time.