Undistillable states cannot yield exact private dits even by successful filtering
OpenQAlgorithms.Papers.Independent.Gap48.P42.no_exact_private_from_undistillableLet be any normalized finite bipartite density operator on positive local dimensions, and suppose its asymptotic Bell-pair distillable entanglement is zero. For every positive integer , every , every pair of positive finite shield dimensions, every normalized shield density , and every family of joint-shield unitaries, form the corresponding exact private state . Then
Each owner retains its own shield, and local ancillas, disposal and arbitrary two-way classical messages are allowed. The conclusion includes deterministic conversion as a special case. It rules out exact finite private targets, while allowing nonzero asymptotic key and approximation by private states with growing shields.
import Definitions.Def_QAlgorithms_Papers_Independent_Gap48_P42
namespace QAlgorithms.Papers.Independent.Gap48.P42
open scoped BigOperators Matrix ComplexConjugate ComplexOrder ENNReal
noncomputable section
attribute [local instance] Classical.propDecidable Classical.decEq
/-- PDF p.4, IV: an undistillable state cannot yield an exact finite private
dit of nontrivial key dimension, even on a positive-probability branch from
any positive finite number of copies. This asserts neither zero asymptotic
key nor a uniform distance from private states with growing shields. -/
theorem no_exact_private_from_undistillable {a b : ℕ} (ha : 0 < a) (hb : 0 < b)
(rho : Matrix (Fin a × Fin b) (Fin a × Fin b) ℂ) (hrho : IsDensityMatrix rho)
(hzero : bipartiteDistillable rho = 0) :
∀ n : ℕ, 0 < n → ∀ d s t : ℕ, 2 ≤ d → 0 < s → 0 < t →
∀ tau : Matrix (Fin s × Fin t) (Fin s × Fin t) ℂ, IsDensityMatrix tau →
∀ U : Fin d → Matrix.unitaryGroup (Fin s × Fin t) ℂ,
¬ CanSelect ![a ^ n, b ^ n] ![d * s, d * t]
(asTwoParties (Dependency50.P42.dimensionedPower rho n))
(asTwoParties (bipartitePrivate tau U)) := by sorry
end
end QAlgorithms.Papers.Independent.Gap48.P42
Read-back
What the Lean code literally says, in plain math · gpt-6.1-sol
Let be positive natural numbers and let be a positive semidefinite trace-one matrix on . Assume that the following supremum in is zero: the supremum of over all nonnegative real such that, for every , all sufficiently large positive integers admit a natural and a density matrix obtainable deterministically from , with and , where is the projector onto . Tensor copies are grouped by party using the fixed cardinality bijections, , and zero target copies are allowed. Then, for every positive integer , every , every positive , every shield density matrix on , and every family of shield unitaries indexed by , no finite adaptive local-operations tree starting from has an output-list entry with . Here has local dimensions , its block between equal row keys and equal column keys is , and all mismatched-key blocks are zero, under the fixed bijections between local key–shield pairs and dimensioned indices; the two-party matrices are represented equivalently as functions on the two local indices. In both the hypothesis’s deterministic conversions and the conclusion’s selectable output, the permitted trees are exactly those built from stopping with the unchanged state, applying an instrument on one party and continuing separately on each outcome, and forgetting an output ensemble by replacing it with its weighted sum when all output local dimensions agree. At an instrument node, branch probabilities are nonnegative and sum to , branch states are density matrices with positive local dimensions, only the acted party’s dimension may change, and lifted local Kraus operators satisfy and ; probabilities multiply along branches. Deterministic obtainability means the final list is , while the forbidden selection requires an exact target entry of any strictly positive probability, with no positive lower bound imposed on that probability.
Confirmed by the mission captain (proposal self-audit).
Confirmed by the moderator at approval.