The von Neumann mean ergodic theorem
ProvedFamousTheorems.von_neumann_mean_ergodicThe von Neumann mean ergodic theorem. Let be a (real or complex) Hilbert space and a bounded linear operator with . Then for every the averages
converge in norm to , where is the orthogonal projection onto the subspace of fixed vectors.
Applied to the Koopman operator of a measure-preserving map , it says that time averages of an function converge in to the conditional expectation onto invariant functions. This is the first ergodic theorem, proved by von Neumann in 1932, and it is basic in ergodic theory and statistical mechanics.
Formalization note. Mathlib's ContinuousLinearMap.tendsto_birkhoffAverage_orthogonalProjection. birkhoffAverage 𝕜 f id n x is . The fixed subspace is f.eqLocus 1, and orthogonalProjectionOnto is the orthogonal projection onto it, coerced into .
import Mathlib
namespace FamousTheorems
theorem von_neumann_mean_ergodic {𝕜 E : Type*} [RCLike 𝕜] [NormedAddCommGroup E] [InnerProductSpace 𝕜 E] [CompleteSpace E]
(f : E →L[𝕜] E) (hf : ‖f‖ ≤ 1) (x : E) :
Filter.Tendsto (fun n => birkhoffAverage 𝕜 f id n x) Filter.atTop
(nhds (((f.eqLocus (1 : E →L[𝕜] E)).orthogonalProjectionOnto x : E))) := by sorry
end FamousTheorems