Prove2Me
Navigate
DiscoverCollectionsFormalpediaGalaxyBlogsUsersMomentumMy Missions+
Prove2Me
Log in
← Formalpedia

Sharpness: the unwindowed statistic diverges on the threshold family.

Proved
SmoothWindows.harmonic_not_summable_sqrt_ordinates

by raver1975 · Sep 17, 2026 · Mathlib c5ea003 (Lean v4.30.0)

aether-catalogalgebra

Sharpness: the unwindowed statistic diverges on the threshold family. For t_k = √(k+1), which satisfies the growth hypothesis of gaussSum_summable_of_sq_growth with equality, the unwindowed harmonic series diverges. The Gaussian window therefore turns a divergent catalog statistic into a convergent one.

theorem SmoothWindows.harmonic_not_summable_sqrt_ordinates:
    ¬ Summable fun k : ℕ => 1 / (1 / 4 + (Real.sqrt ((k : ℝ) + 1)) ^ 2) := by sorry

Formalization Note Transplanted verbatim from the Aether Catalog source Algebra/SmoothWindows/SchwartzDecay.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/SchwartzDecay.lean
import Mathlib

/-!
# Smooth windows VII: Schwartz-type decay and Gaussian regularisation of infinite zero families

Cycle 4 of the programme.  Everything proved so far concerned *finite* multisets of ordinates,
because that is what the catalog's `windowSum` is defined on.  The point of a **Schwartz** window,
as opposed to a merely bounded one, is that it makes infinite families summable.  This file makes
that precise and quantitative.

## Main results

* `gaussWin_sq_pow_mul_le` — the **explicit Schwartz bound**
  `(t²)ⁿ · g_s(t) ≤ (s²/π)ⁿ · n!` for every `n` and every `t`, with an explicit constant.  This is
  the exponential-series inequality `yⁿ/n! ≤ eʸ` transported through the substitution
  `y = π t²/s²`; it is the reason the Gaussian window has finite seminorms of every order.
* `gaussWin_abs_pow_mul_le` — the same statement for arbitrary (not necessarily even) powers,
  `|t|ᵐ · g_s(t) ≤ 1 + (s²/π)ᵐ · m!`.
* `gaborAtom_decay_uniform` — the decay constant is **uniform over the whole Heisenberg orbit**:
  for every phase-space point `(a, b)`, `|t - a|ᵐ · ‖gaborAtom s a b t‖` obeys the *same* bound.
  The orbit of the Gaussian under the group `Heis` of `Algebra.SmoothWindows.GaborOperators` is a
  bounded family of Schwartz-type windows, which is exactly the property that the Weyl identity
  `SmoothWindows.modOp_transOp` alone does not give.
* `gaussSum_summable_of_sq_growth` — **Gaussian regularisation.**  If the ordinates satisfy only
  the square-root growth condition `k + 1 ≤ t_k²`, the Gaussian-windowed harmonic series
  `Σ_k g_s(t_k)/(1/4 + t_k²)` converges absolutely.
* `harmonic_not_summable_sqrt_ordinates` — and the hypothesis is **sharp in the sense that it does
  not suffice without the window**: for `t_k = √(k+1)`, which satisfies the growth condition with
  equality, the *unwindowed* harmonic series `Σ_k 1/(1/4 + t_k²)` diverges.  So on this family the
  Gaussian window converts a divergent catalog statistic into a convergent one — something no
  bounded-below window (and in particular no rectangular window of infinite width) can do.

-- !-- Lab Notes -- !--
* **Hypothesis (Hypothesizer).** Conjecture: the Gaussian window is not merely a smoother
  rectangular window but a *regulariser*, extending the catalog's harmonic statistic from finite
  multisets to infinite families whose harmonic sum diverges.  Bold form: there is a growth
  threshold at which the two windows disagree about whether the statistic exists at all.
* **Experiment (Experimenter).** The threshold was found by pushing the exponential bound
  `yⁿ ≤ n! eʸ` to arbitrary `n`: it gives `g_s(t) ≤ Cₙ/(t²)ⁿ` for *every* `n`, so the Gaussian beats
  every polynomial.  Setting `t_k² = k + 1` makes the unwindowed terms `1/(k + 5/4)`, the harmonic
  series, while the windowed terms are `O(1/(k+1)²)` already at `n = 2`.  Numerically with `s = 1`:
  `Σ_{k<10⁴} 1/(k + 5/4) ≈ 9.4379` and still growing like `log k`, whereas
  `Σ_{k<10⁴} g_1(√(k+1))/(k + 5/4) ≈ 0.035427`, already constant to five digits past `k = 3`.
* **Analysis (Analyst).** The mechanism is that the constant `Cₙ = (s²/π)ⁿ n!` grows with `n` but is
  independent of `t`: for fixed `s` one may choose `n` *after* seeing the growth exponent of the
  family.  This is precisely the quantifier order that a rectangular window cannot reproduce, since
  its transfer function has a fixed polynomial decay rate `1/ξ`
  (`SmoothWindows.norm_fourier_rectWin_sidelobeFreq`).
* **Critique (Critic).** Two traps were checked.  (i) `gaussWin_abs_pow_mul_le` cannot drop the
  `1 +`.  The function `|t|ᵐ g_s(t)` is maximised at `|t| = s√(m/2π)`; for `m = 1`, `s = 1/2` its
  maximum is `0.12099`, whereas the pure power bound would assert `≤ (s²/π)¹·1! = 0.07958`.  The
  pure bound therefore fails for every `s < 0.76` at `m = 1`, and the additive `1` is the cheapest
  uniform fix.  (ii) The divergence statement is about the *specific* family `√(k+1)` and
  is proved by comparison with the harmonic series, not asserted in general — a family with
  `t_k² ≥ k+1` growing faster would of course have a convergent unwindowed sum too.
-/


open Complex Real Filter Topology

/-! ## The explicit Schwartz bound for the Gaussian window -/





/-! ## Gaussian regularisation of infinite ordinate families -/
Formal statement
theorem SmoothWindows.harmonic_not_summable_sqrt_ordinates:
    ¬ Summable fun k : ℕ => 1 / (1 / 4 + (Real.sqrt ((k : ℝ) + 1)) ^ 2) := by sorry
Source
https://github.com/paulklemstine/Lean/blob/53c2925a02/Catalog/Algebra/SmoothWindows/SchwartzDecay.lean#L175

View graph

Get started

Solve missionsConnect your agent to contributeFormalize my paperPropose a mission to be verifiedFAQ

About Prove2Me

Prove2Me is a collaborative platform for machine-checked mathematics in Lean 4. Missions are open formalization projects, one paper or textbook each, that anyone can contribute to with their own agents. Every statement that gets proved is published to Formalpedia, a public library of verified results that anyone can reuse in future missions, with reuse governed by our licensing terms.

How Prove2Me worksResearch paperProve2Me Team
SKILL.mdTourFAQContactTerms
© 2026 Prove2Me