Measure-preserving dynamical system

A measure-preserving dynamical system is a transformation — discrete or continuous — that leaves invariant a measure defined on the underlying space of states. In physical terms: the system evolves, but the total \u0022amount\u0022 of every measurable property remains constant. The most important example in physics is Hamiltonian flow, where Liouville\u0027s theorem guarantees that phase-space volume is preserved.

The concept generalizes beyond mechanics. A baker\u0027s map stretches and folds the unit square in ways that preserve area; the Bernoulli shift permutes sequences in ways that preserve the product measure. These systems are the playground of ergodic theory, which asks not just whether measure is preserved but whether the system eventually distributes that measure uniformly across the space.

Measure preservation is a necessary but not sufficient condition for ergodicity. Many measure-preserving systems are not ergodic: their phase space decomposes into invariant subsets that never mix. The KAM theorem shows that for nearly integrable Hamiltonian systems, most trajectories remain trapped on invariant tori, preserving measure without exploring the full space. Measure preservation guarantees conservation; it does not guarantee exploration.

The significance of measure-preserving dynamics extends to information theory. A measure-preserving transformation is, from the perspective of Kolmogorov-Sinai entropy, a transformation that destroys no information — it merely rearranges it. Whether that rearrangement is computationally simple or effectively random is the question that separates integrable systems from chaotic ones.