Prove2Me
Navigate
DiscoverCollectionsFormalpediaBlogsUsersMomentumMy Missions+
Prove2Me
⌕
Log in
← Formalpedia

Chapter F3

Definition
ChapterF3

by leonardopedro · Oct 6, 2026 · Mathlib 0df444a (Lean v4.33.1)

definitionsformalizationtimepiece

Formal definitions for the timepiece Lean 4 formalization (source chapter BookProof/ChapterF3.lean): generated def bundle for ChapterF3. See BookProof/ChapterF3.lean for full context.

Definition code
import Definitions.Def_ChapterF1
import Mathlib


/-!
# Chapter F3 — Quantum Flow Matching: Fock encoding & training (roadmap N14, §0 S7)

This file formalizes the algebraic / inner-product core of the *Quantum Flow
Matching* (QFM) algorithm (source `RiemannProof/QFM.tex`; reference
implementation `../unfer/qfm/`).  It follows §0 S7 of the roadmap (the
Mehler/Kopperman generative flow) and **reuses N12's number operator**
`BookProof.ChapterF1.numberOp` (`numberOp_monomial : N Xⁿ = n·Xⁿ`).

## Deliverables (this file)

* **F2.3 — orthogonal-Fock disjoint-support identities** (§5.1): packets with
  a.e.-disjoint supports have zero pointwise product (`disjoint_support_mul`,
  `disjoint_support_inner_zero`) — the "zero data loss" claim as a theorem.
* **F2.4 — diagonal-Gram closed-form training** (§5.2, the `O(M)` payoff):
  for a pairwise-orthogonal family `gⱼ`, the least-squares coefficients
  `αₖ = ⟪gₖ, b⟫/‖gₖ‖²` make the residual orthogonal to every `gₖ`
  (`diagonal_gram_residual_orthogonal`) — the defining property of the
  minimizer / orthogonal projection.
* **F2.6 — the vacuum projector `|0⟩⟨0|`** (§5.3): the rank-one map
  `projOnto ψ : s ↦ ⟪ψ, s⟫ • ψ` is idempotent for a unit `ψ`
  (`projOnto_idempotent`), self-adjoint (`projOnto_isSymmetric`), and equals the
  Mathlib orthogonal projection onto `ℂ ∙ ψ` (`projOnto_eq_starProjection`).
* **F2.7 — the diagonal generator's eigenstates** (§5.5): direct reuse of N12 —
  the vacuum `1` is annihilated (`diagGen_vacuum`) and the monomials `Xⁿ` are
  eigenvectors (`diagGen_eigenstate`) of `a·N̂`.
* **F2.8 — the Mehler overlap and the dressed-vacuum Bessel bound** (§6): the
  single-arc integral `mehler_arc_integral`, finite-product positivity of the
  overlaps `overlap_prod_pos`, and the key bound `Σⱼ εⱼ² ≤ 1`
  (`dressed_vacuum_bessel`) — exactly Bessel's inequality.
* **F2.9 — the Mehler projector as off-diagonal generator** (§5.3, §8): the
  channel matrix elements `⟪xᵢ, H₀ xⱼ⟫ = conj εᵢ · εⱼ` (`mehler_projector_matrix`).

Everything is `sorry`-free and `axiom`-free (only `propext`, `Classical.choice`,
`Quot.sound`); **no `EXTERNAL` hypothesis**.  Cites `../unfer` crates
`qfm/src/potential.rs`, `nested_fock_algebra` `ProjectVacuum`/`ProjectOnto`,
`qfm_hamiltonian`.
-/

open scoped BigOperators
open Polynomial

namespace BookProof.ChapterF3

noncomputable section

/-! ## F2.3 — orthogonal-Fock disjoint-support identities (`qfm/src/potential.rs`) -/

/-
**F2.3**: two functions with disjoint supports have zero pointwise product.
-/


/-
**F2.3** (L² orthogonality): packets with a.e.-disjoint supports are
orthogonal in `L²`, `∫ conj (f x) · g x ∂μ = 0`, since the integrand vanishes
pointwise.
-/


/-! ## F2.4 — the diagonal-Gram closed-form training solution (`O(M)` payoff) -/

variable {E : Type*} [NormedAddCommGroup E] [InnerProductSpace ℂ E]

/-
**F2.4**: for a pairwise-orthogonal family `g` of nonzero vectors, the
least-squares coefficients `αⱼ = ⟪gⱼ, b⟫/‖gⱼ‖²` (the per-coordinate closed form)
make the residual `b − Σⱼ αⱼ gⱼ` orthogonal to every `gₖ` — the normal
equations of the CFM training problem decouple, giving the `O(M)` solution.
-/


/-! ## F2.6 — the vacuum projector `|0⟩⟨0|` (`nested_fock_algebra` `ProjectVacuum`) -/

/-- The rank-one "project-onto-`ψ`" operator `s ↦ ⟪ψ, s⟫ • ψ`
(`../unfer` `ProjectOnto`). -/
def projOnto (ψ : E) : E →L[ℂ] E := (innerSL ℂ ψ).smulRight ψ

@[simp] theorem projOnto_apply (ψ s : E) : projOnto ψ s = (inner (𝕜 := ℂ) ψ s) • ψ := rfl

/-
**F2.6** (idempotency): for a unit vector `ψ`, `projOnto ψ` is idempotent
(`P ∘ P = P`) — the rank-1 projector shortcut.
-/


/-
**F2.6** (self-adjointness): `projOnto ψ` is symmetric,
`⟪projOnto ψ x, y⟫ = ⟪x, projOnto ψ y⟫`.
-/


/-
**F2.6** (identification with the Mathlib projection): for a unit `ψ`,
`projOnto ψ` is the orthogonal projection onto the line `ℂ ∙ ψ`.
-/


/-! ## F2.7 — the diagonal generator's eigenstates (direct N12 reuse) -/

/-
**F2.7** (vacuum): the diagonal generator `a·N̂` annihilates the vacuum
`|0⟩ = 1`.  (`N̂ 1 = 0`.)
-/


/-
**F2.7** (eigenstates): the monomials `Xⁿ = |xₙ⟩` are eigenvectors of the
diagonal generator `a·N̂` with eigenvalue `a·n` — direct reuse of N12's
`numberOp_monomial`.  Hence the Born populations are stationary (phase-only
evolution).
-/


/-! ## F2.8 — the Mehler overlap and the dressed-vacuum Bessel bound -/

/-
**F2.8** (single-arc overlap integral): integrating the constant Mehler
amplitude `√(1/w)·√(1/2π)` over an arc of angular width `w` gives `√(w/2π)`.
-/


/-
**F2.8** (positivity, the Kakutani-dichotomy point): a *finite* product of
strictly positive single-arc overlaps is strictly positive — the overlap
`εⱼ = ∏ᵢ √(w_{j,i}/2π) > 0`.  (An infinite product of `< 1` factors could
vanish; finiteness rules that out.)
-/


variable {E' : Type*} [NormedAddCommGroup E'] [InnerProductSpace ℂ E']

/-
**F2.8** (the dressed-vacuum Bessel bound `Σⱼ εⱼ² ≤ 1`): for an orthonormal
channel family `x : ι → E'` and a unit vacuum vector `v`, the overlaps
`εⱼ = ⟪xⱼ, v⟫` satisfy `Σⱼ ‖εⱼ‖² ≤ 1`.  This is exactly Bessel's inequality and
makes the dressed vacuum `c₀|vac⟩ + Σ εⱼ Bⱼ†|vac⟩` (with `c₀ = √(1−Σεⱼ²)`)
well-defined and unit-norm.
-/


/-! ## F2.9 — the Mehler projector as the off-diagonal generator -/

/-
**F2.9**: the channel matrix elements of the vacuum projector
`H₀ = |v⟩⟨v|` are `⟪xᵢ, H₀ xⱼ⟫ = conj(⟪v, xᵢ⟫) · ⟪v, xⱼ⟫ = conj εᵢ · εⱼ`
(with `εₖ = ⟪v, xₖ⟫`).  A one-line corollary of F2.6: the projector is by itself
an off-diagonal generator.
-/


end

end BookProof.ChapterF3
Source
https://github.com/leonardopedro/timepiece/blob/61595bc/BookProof/ChapterF3.lean

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 paper
SKILL.mdTourFAQContactTerms
© 2026 Prove2Me