Recover winding operators from the diagonal blocks and first input column of an actual bounded-operator product
ProvedMasulloRH.cell_operator_product_and_winding_reconstructionIn a complex inner-product space let P be the orthogonal projection onto a subspace admitting one. For 0<r<1, a unit phase u and a sequence F of bounded cell operators, set q=r², a=r conj(u), b=r u. The normalized cell train with coefficients sqrt(1−r²)a^m preserves squared norms and has Gram coefficients (1−q)a^m b^n. Form E(m,n) from these coefficients multiplying I−P, subtracting I only at (0,0). Form the lower triangular product blocks K(i,j) by summing the actual compositions F(i−m) E(m,j) for 0≤m≤i. The theorem recovers F(0) from the diagonal block K(0,0) using an explicit two-sided inverse and recovers every b^j F(j) from two consecutive diagonal blocks, the first-column composition K(j,0)P and the recovered zeroth block. It verifies the special correction at j=1, the sign of the first-column contribution, and the resulting finite sums of all windings. Taking P=0 gives the pole-free case; taking P to be the projection onto the cell pole line gives the principal case.
import Mathlib.Analysis.InnerProductSpace.Projection.Basic import Mathlib.Analysis.SpecificLimits.Normed import Mathlib.Tactic.Abel import Mathlib.Tactic.FieldSimp import Mathlib.Tactic.Linarith import Mathlib.Tactic.NoncommRing import Mathlib.Tactic.Positivity import Mathlib.Tactic.Ring open scoped ComplexConjugate
theorem MasulloRH.cell_operator_product_and_winding_reconstruction {H : Type*} [NormedAddCommGroup H] [InnerProductSpace ℂ H]
(W : Submodule ℂ H) [W.HasOrthogonalProjection]
(F : ℕ → H →L[ℂ] H) (r : ℝ) (hr : 0 < r) (hr' : r < 1)
(u : ℂ) (hu : ‖u‖ = 1) :
let a := (r:ℂ) * (starRingEnd ℂ) u
let b := (r:ℂ) * u
let q := ((r^2 : ℝ) : ℂ)
let train := fun m : ℕ => ((Real.sqrt (1-r^2):ℝ):ℂ)*a^m
let P := W.starProjection
let E := fun m n : ℕ => ((1-q)*a^m*b^n) • (1-P) -
(if m=0 ∧ n=0 then 1 else 0)
let K := fun i j : ℕ => ∑ m ∈ Finset.range (i+1), F (i-m) * E m j
let T := fun j => K j j
let N : H →L[ℂ] H := q • 1 + (1-q) • P
let S : H →L[ℂ] H := q⁻¹ • (1-P) + P
let Z := -(T 0 * S)
let G := fun n : ℕ => (1-q)⁻¹ •
(T (n+1) - q • T n - (if n=0 then q • Z else 0)) -
b^(n+1) • (K (n+1) 0 * P)
(∀ x : H, HasSum (fun m : ℕ => ‖train m • x‖^2) (‖x‖^2)) ∧
(∀ m n, train m * (starRingEnd ℂ) (train n) = (1-q)*a^m*b^n) ∧
N*S=1 ∧ S*N=1 ∧ Z=F 0 ∧
(∀ j, K j 0 * P = -(F j * P)) ∧
(∀ n : ℕ, T (n+1) - q • T n - (if n=0 then q • Z else 0) =
((1-q)*b^(n+1)) • (F (n+1)*(1-P))) ∧
(∀ n : ℕ, G n = b^(n+1) • F (n+1)) ∧
∀ M : ℕ, Z + ∑ n ∈ Finset.range M, G n =
F 0 + ∑ n ∈ Finset.range M, b^(n+1) • F (n+1) := by sorry