§4.3: dephasing separates state and entanglement fidelity
ProvedQAlgorithms.Papers.Independent.Support50.P41.dephasingSeparationLet sum to one, with at least two positive entries, and put . Let be complete dephasing in that same basis, and any normalized purification. Then
This explicit example separates squared state fidelity from entanglement fidelity.
import Definitions.Def_QAlgorithms_Papers_Independent_Support50_P41
noncomputable section
namespace QAlgorithms.Papers.Independent.Support50.P41
open QAlgorithms
open QAlgorithms.Entanglement (IsCompletelyPositive)
open QAlgorithms.Papers.Independent.Dependency50.P40 (IsPurification purificationFidelity)
open QAlgorithms.Papers.Independent.Dependency50.P45 (entropyExchange)
open QAlgorithms.Papers.Independent.Information100.P08 (canonicalPurification)
open QAlgorithms.Papers.Independent.Information100.P34 (choi dephasing)
open QAlgorithms.Papers.Independent.Dependency50.P38 (vectorizeInputFirst)
open QAlgorithms.Papers.Independent.Gap20.P06 (dilationChannel environmentOutput dilatedVector)
open QAlgorithms.Papers.Independent.Information100.P50 (complementaryMap)
open QAlgorithms.Papers.Independent.Renew46.P43 (IsCountablePOVM classicalKLTerm countableEntropy)
open scoped Matrix ComplexConjugate ComplexOrder ENNReal
attribute [local instance] Classical.propDecidable Classical.decEq
-- The accepted generic dilation functions fix classical equality on each factor.
-- Match that instance for Fin before forming their product unitary groups.
local instance (priority := 2000) (n : ℕ) : DecidableEq (Fin n) := Classical.decEq _
/-- PDF p.16: dephasing fixes diagonal inputs but lowers entanglement fidelity for mixed inputs. -/
theorem dephasingSeparation {d a : ℕ} [NeZero d] (p : Fin d → ℝ)
(hp : ∀ i, 0 ≤ p i) (htotal : ∑ i, p i = 1)
(hmixed : ∃ i j, i ≠ j ∧ 0 < p i ∧ 0 < p j)
(psi : EuclideanSpace ℂ (Fin a × Fin d))
(hpsi : IsPurification (Matrix.diagonal fun i => (p i : ℂ)) psi) :
let rho := Matrix.diagonal fun i => (p i : ℂ)
dephasing rho = rho ∧ rootFidelity rho (dephasing rho)^2 = 1 ∧
purificationFidelity psi dephasing = ∑ i, (p i)^2 ∧
purificationFidelity psi dephasing < 1 := by sorry
end QAlgorithms.Papers.Independent.Support50.P41
end
Read-back
What the Lean code literally says, in plain math · gpt-6.1-sol
Let and be integers, and let be nonnegative real numbers with . Suppose there exist distinct indices with and . Put , and let have norm one and reference partial trace . Define by retaining the diagonal entries of and setting every off-diagonal entry to zero. Then
The amplification acts separately on each system block, and the square roots are spectral matrix square roots. The positive-weight hypothesis excludes , and the unit-vector hypothesis excludes ; zero weights at other indices are allowed.
Confirmed by the mission captain (proposal self-audit).
Confirmed by the moderator at approval.