Joint cutoff asymptotic under log-integrability
ProvedMasulloRepair.lower_cutoff_log_asymptoticStatement. Let be a measure space, let be measurable with -almost everywhere, and let be such that both and are -integrable (equivalently: is measurable and ). Then
as and jointly: the difference of the two sides tends to along the product filter . No relation between and is needed.
Where this sits in the manuscript. In the proof of Theorem 12.4.1.1 (S-local trace calculation), equations (12.4.1.23)–(12.4.1.24) on p. 518 reduce the diagonal integral of the kernel to
where the bracket is the logarithmic volume of . Footnote 2 on p. 514 announces that the cutoff is to be replaced by a smoothing of "with not tied to ", and that one must then "study the convergence of the integral in (12.4.1.23), with replaced by , as ". With the lower cutoff the bracket becomes , so the integral is the left-hand side above with
The manuscript states neither an integrability hypothesis on nor a relation between the cutoffs. This theorem is a repair of that step: it gives a precise sufficient condition, namely log-integrability of the integrand together with and (composition of the stated limit with ). The printed case without lower cutoff corresponds to .
Scope. This is a statement about one fixed integrand. It does not treat the sum over (which needs its own uniform tail bound), it does not verify that is integrable in the adelic setting, and it does not establish Theorem 12.4.1.1.
Formalization notes. The complex integrals are Bochner integrals; is the real logarithm, coerced to . The hypotheses are Integrable H ν and Integrable (fun w => H w * log (q w)) ν; the limit is Tendsto … (atTop ×ˢ 𝓝[>] 0) (𝓝 0) for the function of equal to the left side minus the two main terms.
import Mathlib open MeasureTheory Filter Topology
theorem MasulloRepair.lower_cutoff_log_asymptotic
{Ω : Type*} [MeasurableSpace Ω] (ν : Measure Ω) (q : Ω → ℝ) (H : Ω → ℂ)
(hq : Measurable q) (hqpos : ∀ᵐ w ∂ν, 0 < q w)
(hH : Integrable H ν)
(hHlog : Integrable (fun w => H w * (Real.log (q w) : ℂ)) ν) :
Tendsto
(fun p : ℝ × ℝ =>
(∫ w in {w | q w ≤ p.1},
H w * ((Real.log p.1 - Real.log (max (q w) p.2) : ℝ) : ℂ) ∂ν)
- ((Real.log p.1 : ℂ) * ∫ w, H w ∂ν - ∫ w, H w * (Real.log (q w) : ℂ) ∂ν))
(atTop ×ˢ 𝓝[>] (0 : ℝ)) (𝓝 0) := by sorry