GaussWin pos
ProvedSmoothWindows.gaussWin_posaether-catalogalgebra
Formal statement of SmoothWindows.gaussWin_pos from the Aether Catalog (Algebra). The mathematical content is given by the Lean statement below; a human-readable write-up is pending.
theorem SmoothWindows.gaussWin_pos(s t : ℝ) : 0 < gaussWin s t := by sorry
Formalization Note Transplanted verbatim from the Aether Catalog source Algebra/SmoothWindows/GaussianWindow.lean; the statement is byte-identical to the source declaration, elaborated with autoImplicit disabled in the platform environment.
Preamble
-- Thm stub generated from Algebra/SmoothWindows/GaussianWindow.lean
import Mathlib
import Definitions.Def_Algebra_SmoothWindows_GaborOperators
import Definitions.Def_Algebra_SmoothWindows_GaussianWindow
/-!
# Smooth windows II: the Gaussian window and its Fourier duality
Building on the Heisenberg/Weyl algebra of `Algebra.SmoothWindows.GaborOperators`, this file
introduces the **Gaussian window**
`g_s(t) = exp(-π t² / s²)`, `s > 0`,
and establishes the structural facts that make it the canonical smooth replacement for the
rectangular window of `Algebra.ReciprocalZeroHarmonics.WindowDichotomy`.
## Main results
* `gaussWin_translate_mul` — **the Gaussian window algebra is closed under products of
translates**: `g_s(t-a) · g_s(t-b) = exp(-π(a-b)²/(2s²)) · g_{s/√2}(t - (a+b)/2)`. Two Gaussian
probes at different positions multiply to a *single* Gaussian probe at the midpoint, with an
exponentially small overlap constant. Nothing of this kind holds for rectangular windows.
* `fourier_transOp`, `fourier_modOp` — the Fourier transform **intertwines** translation and
modulation: `𝓕 T_a = χ(-a·) 𝓕` and `𝓕 M_b = T_b 𝓕`. Together with the Weyl relation this is
the modulation/translation identity on the frequency side.
* `fourier_gaussC` — **Fourier self-duality with width inversion**: `𝓕 g_s = s · g_{1/s}`. The
Gaussian family is a fixed family of the Fourier transform; the width parameter is inverted,
which is the exact form of the time–frequency uncertainty trade-off for this family.
* `fourier_gaborAtom` — the transform of a Gabor atom `T_a M_b g_s` is a modulated Gaussian
centred at the frequency `b`; `norm_fourier_gaborAtom` shows its modulus is *exactly* a
Gaussian bump, hence **strictly unimodal and sidelobe-free** (`norm_fourier_gaborAtom_lt` and
`norm_fourier_gaborAtom_strictAnti`).
* `gaussWin_tendsto_zero_pow` — the Gaussian window has rapid (super-polynomial) decay, the
Schwartz property that makes it usable as a smooth analysing window.
-- !-- Lab Notes -- !--
* **Hypothesis (Hypothesizer).** Replacing the sharp cutoff by `g_s` should trade a discontinuous
transfer function with slowly decaying sidelobes for a strictly unimodal, exponentially
decaying one, *without* losing any of the exact algebraic identities.
* **Experiment (Experimenter).** The product identity is the polarisation identity
`(t-a)² + (t-b)² = 2(t-m)² + (a-b)²/2` fed through `Real.exp_add`. Self-duality is obtained by
specialising Mathlib's `fourier_gaussian_pi` at `b = 1/s²`, the delicate part being the complex
power `b^{1/2}` which must be identified with the real square root.
* **Analysis (Analyst).** The three phenomena — closure under products, self-duality, unimodality
of the transform — are all consequences of a single fact: the Gaussian is the unique (up to the
Heisenberg action) minimiser of the uncertainty product, hence a *fixed vector* for the
metaplectic action. The rectangular window has none of these properties.
* **Critique (Critic).** Every statement carries `0 < s`; at `s = 0` the definition degenerates
(`x/0 = 0` in Lean gives the constant window `1`), which is why the hypothesis is kept
explicit rather than derived.
-/
open SmoothWindows
open Complex Real MeasureTheory FourierTransform
/-! ## The Gaussian window -/Formal statement
theorem SmoothWindows.gaussWin_pos(s t : ℝ) : 0 < gaussWin s t := by sorry
Source