Chapter HermiteCarlemanEsa
DefinitionChapterHermiteCarlemanEsadefinitionsformalizationtimepiece
Formal definitions for the timepiece Lean 4 formalization (module BookProof.HermiteCarleman, source chapter BookProof/ChapterHermiteCarlemanEsa.lean): Chapter HermiteCarlemanEsa
Generated def bundle for ChapterHermiteCarlemanEsa. See BookProof/ChapterHermiteCarlemanEsa.lean for full context.
Definition code
import Mathlib
import Mathlib
import Definitions.Def_ChapterFarisLavine
import Definitions.Def_ChapterHermiteProductCore
import Definitions.Def_ChapterHermiteRelativeBound
import Definitions.Def_ChapterHyperbolicQuadraticEsa
import Definitions.Def_ChapterStoneBridge
/-!
# A Carleman criterion on the product Hermite basis, and the full diagonal quadratic
family with an arbitrary first-order term
`BookProof.ChapterHermiteRelativeBound` proves that the inhomogeneous quadratic
Hamiltonian
`H = ∑ᵢ cᵢ(πᵢ² + xᵢ²/4) + ∑ᵢ (bᵢxᵢ + b'ᵢπᵢ)`
is essentially self-adjoint on the Gauss–polynomial (product Hermite) core of `L²(ℝᵈ)`
when the quadratic part is **elliptic** (`cᵢ ≥ c₀ > 0`), by a relative bound; the
shifted-core modules (`ChapterShiftedQuadraticEsa`,
`ChapterShiftedQuadraticMatrixEsa`, `ChapterShiftedQuadraticDegenerate`) remove the sign
and the invertibility conditions by completing the square, but need a *classical
equilibrium* — which does not exist in a direction where the quadratic part vanishes and
both `bᵢ` and `b'ᵢ` are present — and they pay for it by moving to a translated,
modulated core. `ChapterQuadratureEsa` settles the opposite extreme, `c = 0`, on the
plain core.
This module removes **all** of those restrictions at once, by a different route: a
*Carleman-type criterion* for the coefficient recursion on the multi-index lattice.
## What is proved
* `LadderRec`, `ladder_eq_zero` — **the instrument.** Let `u : (Fin d →₀ ℕ) → ℂ` be a
square-summable family (only Bessel's inequality `∑_{a ∈ F} ‖u a‖² ≤ B` on finite sets
is used) satisfying, for every multi-index `α`, the nearest-neighbour recursion
`lam α u_α + ∑ᵢ (conj(wᵢ)√(αᵢ+1) u_{α+eᵢ} + wᵢ√αᵢ u_{α−eᵢ}) = z u_α`
with a **real** diagonal `lam` and constant amplitudes `w`, at a point `z` off the real
axis. Then `u = 0`. The proof is the classical Wronskian/flux argument of Carleman,
run on cubes `{α : ∀ i, αᵢ ≤ N}` instead of intervals: the interior contributions are
pairwise conjugate, so the imaginary part of the recursion telescopes to the flux
through the boundary faces (`flux_identity`), which is bounded by `√(N+1)` times the
`ℓ²`-mass carried by those faces (`flux_bound`). The faces are disjoint, so that mass
is summable, while `∑ 1/√(N+1) = ∞` — a contradiction unless the mass vanishes.
* `mixOp_hermiteCore` — the ladder form of `H` on the product Hermite basis: the
quadratic part is diagonal with the real symbol `∑ᵢ cᵢ(αᵢ + ½)`, and the first-order
part raises the `i`-th excitation number with amplitude `wᵢ = bᵢ + ib'ᵢ/2` and lowers
it with `conj wᵢ`.
* `mixOp_deficiencyTrivialAt`, `mixOp_essentiallySelfAdjoint` — **the headline.** For
**arbitrary** real weights `c` (any signs, zeros allowed) and **arbitrary** real
coefficients `b, b'`, the operator `H_c + ∑ᵢ (bᵢxᵢ + b'ᵢπᵢ)` is essentially
self-adjoint on the plain Gauss–polynomial core of `L²(ℝᵈ)`. No ellipticity, no sign
condition, no classical equilibrium, and no change of core.
* `mixOp_stone_flow` — the resulting complete unitary Schrödinger flow, by Stone's
theorem.
* `wave_indefiniteQuadratic_firstOrder_essentiallySelfAdjoint` — the Minkowski corollary:
`□ + V` with `V(t,x) = (t² − ‖x‖²)/4` plus an arbitrary constant external field and an
arbitrary constant boost, on the plain core.
Everything is `sorry`-free and `axiom`-free.
-/
namespace BookProof.HermiteCarleman
open Finset MeasureTheory MvPolynomial
open BookProof.HermiteProductCore BookProof.HermiteProductBasis
open BookProof.FarisLavine
open BookProof.HyperbolicQuadratic
open BookProof.HermiteRelative
open BookProof.StoneBridge BookProof.EsaClosure BookProof.ChapterStoneResolvent
noncomputable section
variable {d : ℕ}
/-! ## 1. The multi-index cube and its faces -/
/-- The cube `{α : ∀ i, αᵢ ≤ N}` of multi-indices, as a finite set. -/
def cube (d N : ℕ) : Finset (Fin d →₀ ℕ) :=
(Fintype.piFinset fun _ : Fin d => Finset.range (N + 1)).image Finsupp.equivFunOnFinite.symm
/-- The interior of the cube in the `i`-th direction: the multi-indices which can still
be raised in that direction without leaving the cube. -/
def inn (d N : ℕ) (i : Fin d) : Finset (Fin d →₀ ℕ) :=
(cube d N).filter (fun a => a i < N)
/-- The `i`-th boundary face of the cube. -/
def face (d N : ℕ) (i : Fin d) : Finset (Fin d →₀ ℕ) :=
(cube d N).filter (fun a => a i = N)
/-! ## 2. The Carleman criterion on the multi-index lattice -/
/-- The raising contribution to the flux at the multi-index `a` in the direction `i`. -/
def rterm (u : (Fin d →₀ ℕ) → ℂ) (amp : Fin d → ℂ) (i : Fin d) (a : Fin d →₀ ℕ) : ℂ :=
(starRingEnd ℂ) (amp i) * ((Real.sqrt ((a i : ℝ) + 1) : ℝ) : ℂ)
* (starRingEnd ℂ) (u a) * u (a + Finsupp.single i 1)
/-- The lowering contribution to the flux at the multi-index `a` in the direction `i`. -/
def lterm (u : (Fin d →₀ ℕ) → ℂ) (amp : Fin d → ℂ) (i : Fin d) (a : Fin d →₀ ℕ) : ℂ :=
amp i * ((Real.sqrt ((a i : ℝ)) : ℝ) : ℂ)
* (starRingEnd ℂ) (u a) * u (a - Finsupp.single i 1)
variable {u : (Fin d →₀ ℕ) → ℂ} {lam : (Fin d →₀ ℕ) → ℝ} {amp : Fin d → ℂ} {z : ℂ}
/-- The nearest-neighbour recursion satisfied by the Hermite coefficients of a deficiency
vector: a real diagonal `lam`, raising amplitude `conj (amp i) √(αᵢ+1)` and lowering
amplitude `amp i √αᵢ`, at the (non-real) point `z`. -/
def LadderRec (u : (Fin d →₀ ℕ) → ℂ) (lam : (Fin d →₀ ℕ) → ℝ) (amp : Fin d → ℂ) (z : ℂ) :
Prop :=
∀ a : Fin d →₀ ℕ,
((lam a : ℝ) : ℂ) * u a
+ ∑ i, ((starRingEnd ℂ) (amp i) * ((Real.sqrt ((a i : ℝ) + 1) : ℝ) : ℂ)
* u (a + Finsupp.single i 1)
+ amp i * ((Real.sqrt ((a i : ℝ)) : ℝ) : ℂ) * u (a - Finsupp.single i 1))
= z * u a
/-! ## 3. The inhomogeneous quadratic Hamiltonian on the Hermite core -/
/-- The inhomogeneous quadratic Hamiltonian `H_c + ∑ᵢ (bᵢxᵢ + b'ᵢπᵢ)` on the
Gauss–polynomial core. -/
def mixOp (c b b' : Fin d → ℝ) : (polyGaussCore (d := d)) →ₗ[ℂ] L2d d :=
quadOp c + foOp b b'
end
end BookProof.HermiteCarleman
Source