The Marcinkiewicz interpolation theorem for sub-additive operators
ProvedDbarRegularity.marcinkiewicz_interpolationLet and let be a map from to measurable functions on . carries Lebesgue measure , and is the norm. Assume that for all :
- for almost every ;
- for and ;
- if : for and ; if : for .
Then for every there is a constant , depending only on , , and , such that for every .
Formalization Note acts on functions EuclideanSpace ℝ (Fin d) → ℂ. Membership in is written out as ∃ f₁ f₂, MemLp f₁ 1 ∧ MemLp f₂ r ∧ f = f₁ + f₂, and measurability of is assumed for such . Sub-additivity is assumed almost everywhere (Stein writes it pointwise). The exponents are in ℝ≥0∞; the case uses Stein's form (17) in place of a weak-type bound. ℝ≥0, and is chosen before and .
import Definitions.Def_DbarRegularity_Setting import Mathlib.Analysis.Calculus.ContDiff.Basic
namespace DbarRegularity
open MeasureTheory Metric Filter Topology
open scoped ENNReal NNReal ContDiff
/-- Stein, *Singular Integrals*, Chapter I, § 4.2, Theorem 5 (p. 21), Marcinkiewicz interpolation:
let `1 < r ≤ ∞` and let `T` map `L¹(ℝ^d) + L^r(ℝ^d)` to measurable functions. Suppose `T` is
sub-additive, of weak type `(1, 1)` with constant `A₁`, and of weak type `(r, r)` with constant
`A_r` (for `r = ∞`: `‖T f‖_∞ ≤ A_r ‖f‖_∞`, Stein's form (17)). Then `‖T f‖_p ≤ A_p ‖f‖_p` for
`f ∈ L^p`, `1 < p < r`, where `A_p` depends only on `A₁`, `A_r`, `p` and `r`. -/
theorem marcinkiewicz_interpolation (r p : ℝ≥0∞) (hr : 1 < r) (hp : 1 < p) (hpr : p < r)
(A₁ Ar : ℝ≥0) :
∃ A : ℝ≥0, ∀ (d : ℕ)
(T : (EuclideanSpace ℝ (Fin d) → ℂ) → EuclideanSpace ℝ (Fin d) → ℂ),
(∀ f, (∃ f₁ f₂, MemLp f₁ 1 volume ∧ MemLp f₂ r volume ∧ f = f₁ + f₂) →
AEStronglyMeasurable (T f) volume) →
(∀ f g, (∃ f₁ f₂, MemLp f₁ 1 volume ∧ MemLp f₂ r volume ∧ f = f₁ + f₂) →
(∃ g₁ g₂, MemLp g₁ 1 volume ∧ MemLp g₂ r volume ∧ g = g₁ + g₂) →
∀ᵐ x, ‖T (f + g) x‖ ≤ ‖T f x‖ + ‖T g x‖) →
(∀ f, MemLp f 1 volume → ∀ α : ℝ≥0, 0 < α →
volume {x | (α : ℝ≥0∞) < ‖T f x‖ₑ} ≤ (A₁ : ℝ≥0∞) / α * eLpNorm f 1 volume) →
(r ≠ ⊤ → ∀ f, MemLp f r volume → ∀ α : ℝ≥0, 0 < α →
volume {x | (α : ℝ≥0∞) < ‖T f x‖ₑ} ≤ ((Ar : ℝ≥0∞) / α * eLpNorm f r volume) ^ r.toReal) →
(r = ⊤ → ∀ f, MemLp f ⊤ volume → eLpNorm (T f) ⊤ volume ≤ Ar * eLpNorm f ⊤ volume) →
∀ f, MemLp f p volume → eLpNorm (T f) p volume ≤ A * eLpNorm f p volume := by sorry
end DbarRegularity
Read-back
What the Lean code literally says, in plain math · claude-opus-5-5
Read-back: DbarRegularity.marcinkiewicz_interpolation
Setting. Exponents with (so , , allowed); constants .
Hypotheses. , , ; real (ℝ≥0).
Definitions used. MemLp, eLpNorm (Lebesgue measure), AEStronglyMeasurable. Write for (pointwise sums of genuine functions).
Conclusion. such that for every and every map satisfying
- is a.e.-strongly measurable for every ;
- for : for a.e. ;
- for , : ;
- if : for , : ;
- if : for : ;
one has for all .
Quantifier order: , . So is independent of and .
Degenerate cases. : hypothesis 4 is vacuous and 5 is active; when , 5 is vacuous. forces a.e. on ; fine. is allowed (one-point probability space); the Marcinkiewicz theorem holds on any measure space, so no problem.
Remarks. acts on functions, not a.e.-classes; hypotheses and conclusion are all about the same representatives, which is enough for the standard proof (split , the pieces are in and respectively, apply 2 for each fixed , integrate the distribution function). No homogeneity of is assumed, and none is needed.
Soundness checks.
ENNReal: so ; under theMemLphypotheses, so no issue. In 4, and , sor.toReal(thetoReal ⊤ = 0trap is avoided by the guardr ≠ ⊤). isℝ≥0, so finite. No Bochner integrals. Hypotheses are satisfiable (e.g. , or = identity); the conclusion is non-trivial.- is chosen before and : correct and true, since the Marcinkiewicz constant depends only on (Stein Ch. I, Thm 5). Plausibly TRUE as written.
Verdict: no mis-formalization observed
Confirmed by the mission captain (proposal self-audit).
Confirmed by the moderator at approval.