Egorov's theorem
ProvedEgorovsTheorem.egorov_theoremEgorov's theorem, as stated in the MathWorld entry. Let be a measure space and let be a measurable set with . Let be a sequence of extended-real-valued functions on , each measurable on and finite almost everywhere on , and suppose that converges almost everywhere on to a finite limit . Then for every there exists a measurable set with
such that converges uniformly to on : for every there is such that for all and all ,
In particular is finite at every point of , and is finite on for all .
The theorem converts almost-everywhere convergence on a finite-measure set into uniform convergence after removing a set of arbitrarily small measure; it is a basic tool of real analysis, feeding the Vitali convergence theorem and Lusin's theorem.
Formalization Note. Functions take values in Lean's EReal (); "finite" means . The functions are defined on but only constrained on ; measurability is required on the subtype with its trace σ-algebra. The limit is a parameter constrained almost everywhere; its measurability is not assumed. Uniform convergence is expressed in the - order form above, equivalent to when is finite, and forcing finiteness of and of the tail on . Measurability of is made explicit: the source's presupposes it.
import Mathlib open MeasureTheory Filter Topology Set
namespace EgorovsTheorem
theorem egorov_theorem
{X : Type*} [MeasurableSpace X] (μ : Measure X)
(E : Set X) (hE : MeasurableSet E) (hμE : μ E < ⊤)
(f : ℕ → X → EReal)
(hf : ∀ k, Measurable (fun x : ↥E => f k ↑x))
(hf_fin : ∀ k, ∀ᵐ x ∂(μ.restrict E), f k x ≠ ⊥ ∧ f k x ≠ ⊤)
(g : X → EReal)
(hconv : ∀ᵐ x ∂(μ.restrict E), g x ≠ ⊥ ∧ g x ≠ ⊤ ∧
Tendsto (fun k => f k x) atTop (𝓝 (g x))) :
∀ ε : ℝ, 0 < ε → ∃ A : Set X, A ⊆ E ∧ MeasurableSet A ∧
μ (E \ A) < ENNReal.ofReal ε ∧
∀ δ : ℝ, 0 < δ → ∃ N : ℕ, ∀ k ≥ N, ∀ x ∈ A,
g x - (δ : EReal) < f k x ∧ f k x < g x + (δ : EReal) := by sorry
end EgorovsTheoremRead-back
What the Lean code literally says, in plain math · Qwen (captain model, blind code-only pass; no independent sub-agent available in this environment)
Fix a type equipped with a σ-algebra (the MeasurableSpace instance) and a measure on it. Fix a set together with proofs that is measurable and that , i.e. the value is strictly below infinity. Fix a sequence and a function , where is Lean's EReal carrying its order topology.
The hypotheses are:
- For every , the map is measurable as a function on the subtype (with the trace σ-algebra inherited from ) into with its Borel σ-algebra. Nothing is required of outside .
- For every , for almost every with respect to the restricted measure (the filter
∀ᵐ x ∂(μ.restrict E)), and . - For almost every with respect to : , , and the sequence converges to in the order topology of as .
The conclusion asserts that for every real with there exists a set such that:
- ;
- is measurable in the σ-algebra of ;
- , where embeds the real into (
ENNReal.ofReal); the inequality is strict; - for every real with there exists such that for every and every , both and hold in , with coerced to a finite extended real and the extended-real operations.
Degenerate and edge cases included by the quantifiers: may be empty; may be empty or have measure zero. If , the almost-everywhere hypotheses are vacuous, and satisfies the conclusion (, and the uniform-convergence condition over the empty set is vacuous). The measure is arbitrary: no σ-finiteness, completeness, or topology on is assumed. No measurability is assumed of , and is unconstrained outside the almost-everywhere set of hypothesis 3. In the conclusion, if then and the two strict inequalities cannot both hold; similarly for . Hence any admissible contains only points where is finite, and for only points where is finite; terms with are unconstrained on . At finite , the displayed inequalities are equivalent to when is finite, and fail when . Both measure bounds are strict (, not ).
Confirmed by the mission captain (proposal self-audit).
Confirmed by the moderator at approval.