Translation becomes modulation.
ProvedSmoothWindows.fourier_transOpaether-catalogalgebra
Translation becomes modulation. π (T_a f)(w) = Ο(-aw) π f (w).
theorem SmoothWindows.fourier_transOp(a : β) (f : β β β) (w : β) :
π (transOp a f) w = chi (-(a * w)) * π f w := 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(-Ο tΒ² / sΒ²)`, `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` β **the Gaussian window algebra is closed under products of
translates**: `g_s(t-a) Β· g_s(t-b) = exp(-Ο(a-b)Β²/(2sΒ²)) Β· g_{s/β2}(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` β the Fourier transform **intertwines** translation and
modulation: `π T_a = Ο(-aΒ·) π` and `π M_b = T_b π`. Together with the Weyl relation this is
the modulation/translation identity on the frequency side.
* `fourier_gaussC` β **Fourier self-duality with width inversion**: `π g_s = s Β· 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 timeβfrequency uncertainty trade-off for this family.
* `fourier_gaborAtom` β 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` β 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)Β² + (t-b)Β² = 2(t-m)Β² + (a-b)Β²/2` fed through `Real.exp_add`. Self-duality is obtained by
specialising Mathlib's `fourier_gaussian_pi` at `b = 1/sΒ²`, the delicate part being the complex
power `b^{1/2}` which must be identified with the real square root.
* **Analysis (Analyst).** The three phenomena β closure under products, self-duality, unimodality
of the transform β 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 -/
/-! ## Fourier intertwining of translation and modulation -/Formal statement
theorem SmoothWindows.fourier_transOp(a : β) (f : β β β) (w : β) :
π (transOp a f) w = chi (-(a * w)) * π f w := by sorrySource