Two-step Carleman criterion: lattice fluxes, recursion and the criterion
DefinitionChapterCarlemanTwoStepPromotedFormal definitions for the timepiece Lean 4 formalization (source chapter BookProof/ChapterCarlemanTwoStep.lean): the two-step Carleman criterion on the multi-index lattice. The platform's served Def_ChapterCarlemanTwoStep copy is a skeleton (a bare noncomputable section with no namespace), so every consumer statement that says open BookProof.CarlemanTwoStep dies with unknown namespace, and published bundles are immutable, so the fix is a new bundle name that keeps the original namespace (the new-name route). This bundle carries the chapter's definitions: the interior and k-thick boundary faces innK/faceK of a lattice cube, the single-hop contributions rtermG/ltermG, the one- and two-step amplitudes rc1/lc1/rc2/lc2, and the two-step recursion LadderRec2. The chapter's theorem facts stay problem nodes. Consumers add a single import Definitions.Def_ChapterCarlemanTwoStepPromoted.
-- §2.13 new-name route (Session 40): the platform's served `Def_ChapterCarlemanTwoStep` copy is a
-- skeleton (no declarations at all), so `open BookProof.CarlemanTwoStep` in every consumer statement
-- dies with `unknown namespace`. `submit-definition` cannot update an existing
-- name, so this bundle carries the chapter's DEFINITIONS in the original
-- namespace; its theorem facts stay problem nodes. Consumers add a single
-- `import Definitions.Def_ChapterCarlemanTwoStepPromoted`.
-- Left to the problem nodes on purpose (theorem facts):
-- sub_add_singleK
-- sub_singleK_apply
-- rc1_nonneg
-- lc2_vanish
import Definitions.Def_ChapterHermiteCarlemanEsaPromoted
import Mathlib
/-!
# A two-step Carleman criterion on the multi-index lattice
`BookProof.ChapterHermiteCarlemanEsa` proves a Carleman criterion for a *nearest
neighbour* recursion on the lattice of multi-indices: the hops are `α ↦ α ± eᵢ`, with
amplitudes of size `O(√αᵢ)`. That is exactly the ladder structure of a **diagonal**
quadratic Hamiltonian `∑ᵢ cᵢ(πᵢ² + xᵢ²/4)` plus a first-order term.
A general **mode-diagonal** quadratic Hamiltonian
`H = ∑ᵢ (pᵢπᵢ² + qᵢxᵢ² + sᵢ·½(xᵢπᵢ + πᵢxᵢ)) + ∑ᵢ (bᵢxᵢ + b'ᵢπᵢ)`
is not of that form: `xᵢ²`, `πᵢ²` and the squeezing generator `½(xᵢπᵢ + πᵢxᵢ)` all
contain `aᵢ†²` and `aᵢ²`, which move the `i`-th excitation number by **two**, with an
amplitude of size `O(αᵢ)`. This module proves the Carleman criterion for such a
recursion: hops `α ↦ α ± eᵢ` *and* `α ↦ α ± 2eᵢ`, with amplitudes `O(N)` on the boundary
of the cube `{α : ∀ i, αᵢ ≤ N}`.
## What is proved
* `innK`, `faceK` — the interior and the `k`-thick boundary face of a cube in a fixed
direction; `sum_shiftK`, `sum_cube_splitK` — the reindexing and splitting identities.
* `rtermG`, `ltermG`, `sum_cube_hop_im` — **the abstract flux cancellation**: for a
single Hermitian hop family of step `k`, the interior contributions occur in conjugate
pairs, so the imaginary part of the total contribution over a cube is carried entirely
by the `k`-thick boundary face.
* `LadderRec2`, `flux_identity2` — the two-step recursion and its flux identity.
* `flux_boundG` — the flux through a face is at most the amplitude bound there times the
`ℓ²`-mass carried by the face and its shift.
* `sum_range_of_multiplicity`, `faceK_multiplicity`, `shiftedK_multiplicity` — Bessel's
inequality with multiplicity: a `k`-thick face meets at most `k` cubes, so the total
face mass is at most `k` times the total mass. (For `k = 1` the faces are disjoint;
for `k = 2` they are not, and this is what replaces disjointness.)
* `ladder2_eq_zero` — **the criterion.** A square-summable family satisfying the
two-step recursion with a real diagonal and constant amplitudes, at a point off the
real axis, vanishes. The Carleman divergence used is `∑ 1/(N+1) = ∞`.
Everything is `sorry`-free and `axiom`-free.
-/
namespace BookProof.CarlemanTwoStep
open Finset
open BookProof.HermiteCarleman
noncomputable section
variable {d : ℕ}
/-! ## 1. Cubes with a step -/
/-- The part of the cube that can still be raised by `k` in the direction `i` without
leaving the cube. -/
def innK (d N : ℕ) (i : Fin d) (k : ℕ) : Finset (Fin d →₀ ℕ) :=
(cube d N).filter (fun a => a i + k ≤ N)
/-- The `k`-thick boundary face of the cube in the direction `i`. -/
def faceK (d N : ℕ) (i : Fin d) (k : ℕ) : Finset (Fin d →₀ ℕ) :=
(cube d N).filter (fun a => N < a i + k)
/-! ## 2. The abstract flux cancellation for one hop family -/
variable {u : (Fin d →₀ ℕ) → ℂ}
/-- The raising contribution of a hop of step `k` in the direction `i`. -/
def rtermG (u : (Fin d →₀ ℕ) → ℂ) (w : ℂ) (rc : (Fin d →₀ ℕ) → Fin d → ℝ) (k : ℕ) (i : Fin d)
(a : Fin d →₀ ℕ) : ℂ :=
(starRingEnd ℂ) w * ((rc a i : ℝ) : ℂ) * (starRingEnd ℂ) (u a) * u (a + Finsupp.single i k)
/-- The lowering contribution of a hop of step `k` in the direction `i`. -/
def ltermG (u : (Fin d →₀ ℕ) → ℂ) (w : ℂ) (lc : (Fin d →₀ ℕ) → Fin d → ℝ) (k : ℕ) (i : Fin d)
(a : Fin d →₀ ℕ) : ℂ :=
w * ((lc a i : ℝ) : ℂ) * (starRingEnd ℂ) (u a) * u (a - Finsupp.single i k)
/-! ## 3. The two-step recursion -/
/-- The raising coefficient of a one-step hop: `√(αᵢ+1)`. -/
def rc1 (a : Fin d →₀ ℕ) (i : Fin d) : ℝ := Real.sqrt ((a i : ℝ) + 1)
/-- The lowering coefficient of a one-step hop: `√αᵢ`. -/
def lc1 (a : Fin d →₀ ℕ) (i : Fin d) : ℝ := Real.sqrt (a i : ℝ)
/-- The raising coefficient of a two-step hop: `√((αᵢ+1)(αᵢ+2))`. -/
def rc2 (a : Fin d →₀ ℕ) (i : Fin d) : ℝ := Real.sqrt (((a i : ℝ) + 1) * ((a i : ℝ) + 2))
/-- The lowering coefficient of a two-step hop: `√(αᵢ(αᵢ−1))`. -/
def lc2 (a : Fin d →₀ ℕ) (i : Fin d) : ℝ := Real.sqrt ((a i : ℝ) * ((a i : ℝ) - 1))
/-- **The two-step recursion.** A real diagonal `lam`, one-step amplitudes `w1` and
two-step amplitudes `w2`, at the point `z`. -/
def LadderRec2 (u : (Fin d →₀ ℕ) → ℂ) (lam : (Fin d →₀ ℕ) → ℝ) (w1 w2 : Fin d → ℂ) (z : ℂ) :
Prop :=
∀ a : Fin d →₀ ℕ,
((lam a : ℝ) : ℂ) * u a
+ ∑ i, ((starRingEnd ℂ) (w1 i) * ((rc1 a i : ℝ) : ℂ) * u (a + Finsupp.single i 1)
+ w1 i * ((lc1 a i : ℝ) : ℂ) * u (a - Finsupp.single i 1))
+ ∑ i, ((starRingEnd ℂ) (w2 i) * ((rc2 a i : ℝ) : ℂ) * u (a + Finsupp.single i 2)
+ w2 i * ((lc2 a i : ℝ) : ℂ) * u (a - Finsupp.single i 2))
= z * u a
variable {lam : (Fin d →₀ ℕ) → ℝ} {w1 w2 : Fin d → ℂ} {z : ℂ}
/-! ## 4. The flux bound -/
/-! ## 5. Bessel's inequality with multiplicity -/
/-! ## 6. The criterion -/
end
end BookProof.CarlemanTwoStep