Unital qubit channels: norm multiplicativity and entropy and Holevo additivity
OpenQAlgorithms.Papers.Independent.Support50.P13.unitalQubitAdditivityLet be a completely positive trace-preserving linear map satisfying . Let be an integer and let be any completely positive trace-preserving linear map. A density matrix is positive semidefinite with trace one. For a positive matrix and a finite real , put and , where ranges over every input density matrix. Write , with natural logarithms and , and . The Holevo quantity is the supremum of over all finite ensembles of density matrices with nonnegative weights summing to one. Then
The product-channel optimizations include all entangled input density matrices. These equalities assert that entanglement between the two channel inputs does not improve these three optimized quantities.
import Definitions.Def_QAlgorithms_Papers_Independent_Support50_P13
noncomputable section
open scoped Matrix ComplexConjugate ComplexOrder
open Filter
namespace QAlgorithms.Papers.Independent.Support50.P13
open QAlgorithms.Papers.Independent.Gap48.P03
(productChannel maximumOutputPNorm holevoQuantity depolarizingNormFactor
phaseChannel qubitPauli)
open QAlgorithms.Papers.Independent.Next46.P42 (minimumOutputEntropy)
open QAlgorithms.Papers.Independent.Information100.P63 (schatten)
open QAlgorithms.Papers.Independent.Dependency50.P23 (ordinaryClassicalCapacity)
attribute [local instance] Classical.propDecidable Classical.decEq
/-- King quant-ph/0103156v2, PDF p.4, Theorem 1, (5)-(7).
Phi is an arbitrary unital qubit channel, with no standard-form restriction.
Omega is an arbitrary finite positive-dimensional CPTP endomap.
The imported minimumOutputEntropy takes values in the extended nonnegative reals;
on these finite-dimensional channels its values are finite and measured in nats. -/
theorem unitalQubitAdditivity
(Phi : Superop (Fin 2) (Fin 2)) (hPhi : IsChannel Phi)
(hPhiUnital : Phi 1 = 1)
(K : ℕ) (hK : 0 < K)
(Omega : Superop (Fin K) (Fin K)) (hOmega : IsChannel Omega) :
(∀ p : ℝ, 1 ≤ p →
maximumOutputPNorm (productChannel Omega Phi) p =
maximumOutputPNorm Omega p * maximumOutputPNorm Phi p) ∧
minimumOutputEntropy (productChannel Omega Phi) =
minimumOutputEntropy Omega + minimumOutputEntropy Phi ∧
holevoQuantity (productChannel Omega Phi) =
holevoQuantity Omega + holevoQuantity Phi := by sorry
end QAlgorithms.Papers.Independent.Support50.P13
end
Read-back
What the Lean code literally says, in plain math · gpt-6.1-sol
Let be a channel satisfying , let be a positive natural number, and let be a channel. Here being a channel means admitting a finite representation with . Define their product by
For a matrix , write , where lists its Hermitian eigenvalues in decreasing order with multiplicity; the underlying eigenvalue definition instead returns zeros when is not Hermitian. Put , with ranging over positive semidefinite matrices of trace , and put
Finally, define , where
ranges over all finite subsets of the output index set, including the empty subset, is the corresponding principal submatrix, and matrix functions use continuous functional calculus. The infimum is over the coefficient states defined by a summable nuclear expansion, nonnegative finite quadratic forms, and summable real diagonal of sum , which in these finite dimensions are density matrices; its values lie in . Then all three assertions hold:
The entropy-minimum equality is an equality of extended nonnegative real numbers, including possible infinite values.
Confirmed by the mission captain (proposal self-audit).
Confirmed by the moderator at approval.