Truncated rescaled saddle integrands
DefinitionAnalyticCombinatorics_Ch8_SaddlePoint_GaussianCoreanalytic-combinatoricslean4saddle-point
Hfun evaluates the normalized saddle integrand at angle x/sqrt(B_n) inside the window |x|≤delta_n sqrt(B_n), and is zero outside. This is the actual truncated function used for Gaussian dominated-convergence estimates.
Definition code
import Definitions.Def_AnalyticCombinatorics_Ch8_SaddlePoint_Assembly
import Mathlib
import Mathlib.Analysis.Analytic.Basic
import Mathlib.Analysis.Analytic.OfScalars
import Mathlib.Analysis.Analytic.RadiusLiminf
import Mathlib.RingTheory.PowerSeries.Basic
import Mathlib.Algebra.Algebra.Rat
import Mathlib.Data.Complex.Basic
set_option autoImplicit true
set_option relaxedAutoImplicit false
set_option maxSynthPendingDepth 3
/-!
# Gaussian central saddle core
This file packages the reusable dominated-convergence core for the central
Gaussian window after the saddle scaling `theta = x / sqrt (B n)`.
-/
open Complex Filter Asymptotics MeasureTheory
open scoped Topology Real NNReal ENNReal Interval
noncomputable section
/-- The scaled central integrand with the moving cutoff built in. -/
def Hfun (f : ℂ → ℂ) (r B δ : ℕ → ℝ) (n : ℕ) (x : ℝ) : ℂ :=
if |x| ≤ δ n * Real.sqrt (B n) then
saddleG f r n (x / Real.sqrt (B n))
else
0
/-
Note: the prompt-shaped version dropping the measurability hypothesis is omitted;
Mathlib's dominated convergence genuinely requires `AEStronglyMeasurable` of the
integrands, which is not implied by pointwise convergence + domination for an
arbitrary `f`. The clean theorem `central_tendsto_one_of_dominated_of_aestronglyMeasurable`
takes it as a hypothesis (discharged easily for continuous integrands, e.g. the exp testbed).
-/
Source