Proposition II.1: finite-block decorrelation converse
OpenQAlgorithms.Papers.Independent.Intakenext100.P47.decorrelation_converseLet be a normalized bipartite density matrix on finite nonzero spaces of dimensions , put , and let . Suppose a coordinated local product-unitary ensemble , with probabilities and terms, -decorrelates for : its output is within full trace norm of some normalized product state. Define for , with , and for . Then
Here is the natural exponential constant. The factor multiplies the entire parenthesis. This is the finite-block converse for total correlation erasure, including large error tolerances.
import Definitions.Def_QAlgorithms_Papers_Independent_Intakenext100_P47
noncomputable section
namespace QAlgorithms.Papers.Independent.Intakenext100.P47
open QAlgorithms Filter
open QAlgorithms.Papers.Independent.Gap48.P20 (Copies localProduct)
open QAlgorithms.Papers.Independent.Dependency50.P19 (mutualInformation first)
open QAlgorithms.Papers.Independent.Dependency50.P45 (entropyExchange)
open QAlgorithms.Papers.Independent.Information100.P27 (shannon Config)
open QAlgorithms.Papers.Independent.Information100.P71 (channelPower)
open QAlgorithms.Papers.Independent.Advance50.P05 (RectangularInstrument)
open scoped Matrix ComplexConjugate
attribute [local instance] Classical.propDecidable Classical.decEq
/-- PDF p.3, Proposition II.1: the outer n multiplies erasureEta(3 eps) as printed. -/
theorem decorrelation_converse {A B : Type} [Fintype A] [Fintype B]
[Nonempty A] [Nonempty B] (rho : Matrix (A × B) (A × B) ℂ)
(hrho : IsDensityMatrix rho) (n : ℕ) (hn : 0 < n) (eps : ℝ) (heps : 0 ≤ eps)
(R : COLUR (Copies A n) (Copies B n))
(hR : Decorrelates (R.operation (sourcePower rho n)) eps) :
let lower := (n : ℝ) * (mutualInformation rho - 3 * eps * logDimension A B - erasureEta (3*eps))
lower ≤ entropyExchange R.operation (sourcePower rho n) ∧
lower ≤ shannon R.weight ∧ lower ≤ Real.logb 2 R.count := by sorry
end QAlgorithms.Papers.Independent.Intakenext100.P47
end
Read-back
What the Lean code literally says, in plain math · gpt-6.1-sol
For any finite nonempty types , any positive semidefinite, trace-one matrix on , any positive natural number , any real , and any positive-length list of pairs of local unitaries on , with nonnegative weights summing to , define and . Assume that there exist positive semidefinite, trace-one matrices such that , where . Put , , where and , and define for and otherwise. Then each of
is at least , where and matrix functions use continuous functional calculus. The tolerance may be or arbitrarily large, the displayed lower bound need not be positive, and zero weights are allowed.
Confirmed by the mission captain (proposal self-audit).
Confirmed by the moderator at approval.