First saddle correction polynomial and absolute scale
DefinitionAnalyticCombinatorics_Ch8_SaddlePoint_SecondOrderCoreanalytic-combinatoricslean4saddle-point
The complex polynomial P2 contains cubic imaginary, quartic and sextic terms. Gaussian integration gives d1=b4/(8b²)-5b3²/(24b³), while, for b>0, the nonnegative comparison expression is Q2=|b4|/b²+b3²/b³.
Definition code
import Mathlib.Probability.Distributions.Gaussian.Real
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
/-!
# Second-order saddle core definitions
This file defines the universal Gaussian correction polynomial and the scalar
second-order correction appearing after integrating its even moments.
-/
open Complex Filter Asymptotics MeasureTheory
open scoped Topology Real NNReal ENNReal Interval
noncomputable section
/-- Second-order local polynomial in the scaled variable `x`. -/
def saddleSecondPoly (b b3 b4 : ℝ → ℝ) (r x : ℝ) : ℂ :=
1
- Complex.I * ((b3 r : ℂ) / (6 * (Real.sqrt (b r) : ℂ) ^ 3)) * (x : ℂ) ^ 3
+ ((b4 r : ℂ) / (24 * (b r : ℂ) ^ 2)) * (x : ℂ) ^ 4
- (((b3 r) ^ 2 : ℝ) : ℂ) / (72 * (b r : ℂ) ^ 3) * (x : ℂ) ^ 6
/-- The scalar second-order saddle correction. -/
def saddleCorrection (b b3 b4 : ℝ → ℝ) (r : ℝ) : ℝ :=
b4 r / (8 * (b r) ^ 2) - 5 * (b3 r) ^ 2 / (24 * (b r) ^ 3)
/--
Robust absolute correction scale. This is the scale used before specializing
to the signed correction, so cancellation in `saddleCorrection` does not hide
the true error size.
-/
def saddleCorrectionScale (b b3 b4 : ℝ → ℝ) (r : ℝ) : ℝ :=
|b4 r| / (b r) ^ 2 + (b3 r) ^ 2 / (b r) ^ 3
@[simp] lemma saddleSecondPoly_apply (b b3 b4 : ℝ → ℝ) (r x : ℝ) :
saddleSecondPoly b b3 b4 r x =
1
- Complex.I * ((b3 r : ℂ) / (6 * (Real.sqrt (b r) : ℂ) ^ 3)) * (x : ℂ) ^ 3
+ ((b4 r : ℂ) / (24 * (b r : ℂ) ^ 2)) * (x : ℂ) ^ 4
- (((b3 r) ^ 2 : ℝ) : ℂ) / (72 * (b r : ℂ) ^ 3) * (x : ℂ) ^ 6 := rfl
@[simp] lemma saddleCorrection_apply (b b3 b4 : ℝ → ℝ) (r : ℝ) :
saddleCorrection b b3 b4 r =
b4 r / (8 * (b r) ^ 2) - 5 * (b3 r) ^ 2 / (24 * (b r) ^ 3) := rfl
@[simp] lemma saddleCorrectionScale_apply (b b3 b4 : ℝ → ℝ) (r : ℝ) :
saddleCorrectionScale b b3 b4 r =
|b4 r| / (b r) ^ 2 + (b3 r) ^ 2 / (b r) ^ 3 := rfl
Source