Maximal ergodic theorem. #
In this file we prove two closely related versions of maximal ergodic theorem for a
measure-preserving map f.
The first version deals with integrable real-valued functions g, and states that the integral of
g over the set $\left\{ x : \sup_{n ≥ 0} \sum_{i = 0}^{n - 1} (g ∘ f ^ i)(x) > 0\right\}$ is
nonnegative.
The second version, a corollary of the first, states that for integrable normed-space valued g,
the operator $g ↦ \sup_{n ≥ 0} \frac{1}{n}\sum_{i=0}^{n-1}g ∘ f^i$ satisfies a weak-type
inequality. In other words, its cumulative distribution function
a ↦ a * μ.real {x | a < birkhoffAverageSup f g x} is bounded above by the L1 norm of g.
Main definitions #
birkhoffMax f g n: the maximum ofbirkhoffSum f g iforiranging from0ton.birkhoffSumSup f g: the supremum of the Birkhoff sums ofgalong orbits off.birkhoffAverageSup f g: the maximal ergodic operator, defined as the supremum of the Birkhoff averages ofgalong orbits off.
Main results #
setIntegral_birkhoffSumSup_nonneg: for a measure-preservingf, the integral of an integrable real-valuedgover the set wherebirkhoffSumSup f gis positive is non-negative.mul_measureReal_lt_birkhoffAverageSup_le_integral: the cumulative distribution function ofbirkhoffAverageSupatais less than or equal to the integral ofgon the set wherea < birkhoffAverageSup f g x.mul_measureReal_lt_birkhoffAverageSup_le_norm: the operatorbirkhoffAverageSupsatisfies a weak-type inequality.
Implementation #
Write $M_n$ for birkhoffMax f g n, which is non-negative and monotone in n. We have a
recurrence relation $M_{n+1} = g + M_n ∘ f$ in the set where $M_{n+1} > 0$ (birkhoffMax_succ'),
so by monotonicity of $M_n$, $M_n ≤ 1_{\{M_n > 0\}} g + M_n ∘ f$ pointwise
(birkhoffMax_le_indicator_add_comp). Integrating this inequality and using
that f preserves μ, we conclude that $\int_{\{M_n > 0\}} g$ is nonnegative
(setIntegral_support_birkhoffMax_nonneg). The sets $\{M_n > 0\}$ increase to the set where
birkhoffSumSup f g is positive (setOf_birkhoffSumSup_pos_eq_iUnion_birkhoffMax_support), so
setIntegral_birkhoffSumSup_nonneg follows by passing to the limit
(tendsto_setIntegral_support_birkhoffMax).
For 0 ≤ a, the set where a < birkhoffAverageSup f g equals the set where
birkhoffSumSup f (g - a) is positive (lt_birkhoffAverageSup_iff_lt_birkhoffSumSup). Applying
the previous result to g - a on this set gives
mul_measureReal_lt_birkhoffAverageSup_le_integral. Applying that to ‖g ·‖ gives the
weak-type inequality.
The maximum of the Birkhoff sums of g along orbits of f from 0 to n.
Equations
- birkhoffMax f g = partialSups (birkhoffSum f g)
Instances For
The supremum of the Birkhoff sums of g along orbits of f.
Equations
- birkhoffSumSup f g x = ⨆ (n : ℕ), ↑(birkhoffSum f g n x)
Instances For
The maximal ergodic operator: the supremum of the Birkhoff averages of g.
Equations
- birkhoffAverageSup f g x = ⨆ (n : ℕ), ↑(birkhoffAverage ℝ f g n x)
Instances For
The integral of g over the set where birkhoffSumSup f g is positive is non-negative.
The cumulative distribution function of birkhoffAverageSup at a is less than or equal to the
integral of g on the set where a < birkhoffAverageSup f g x.
Maximal ergodic theorem: the operator birkhoffAverageSup satisfies a weak-type inequality.