A.2: coherent-information subadditivity and channel additivity (72)
OpenQAlgorithms.Papers.Independent.Gap20.P19.coherentInformationAdditiveAll Hilbert spaces are finite dimensional, complex, and nonzero. A channel is completely positive and trace preserving on every operator. Write for its Choi rank, the minimum pure-environment dimension. Receiver and complementary channels are always obtained from the same isometry.
Let be isometric dilations of two degradable channels . Write and over all density matrices. For every joint input state , with its computed marginals ,
Moreover and for every nonnegative integer , with the scalar empty tensor at .
This is information additivity, without asserting an additional operational coding theorem.
import Definitions.Def_QAlgorithms_Papers_Independent_Gap20_P19
noncomputable section
namespace QAlgorithms.Papers.Independent.Gap20.P19
open scoped BigOperators Matrix ComplexConjugate ComplexOrder Kronecker
open Filter Topology
open QAlgorithms.Papers.Independent.Information100.P67
(IsIsometry receiverChannel complementaryChannel IsDegradable IsAntidegradable
maximumCoherentInformation dilationPower)
open QAlgorithms.Papers.Independent.Claimed50.P48
(MinimalDilation indexedKrausDilation arbitraryVectorComplement superopTensor tensorDilation)
open QAlgorithms.Papers.Independent.Claimed50.P24 (ChannelEB)
open QAlgorithms.Papers.Independent.Information100.P60 (wernerHolevo)
open QAlgorithms.Papers.Independent.Gap48.P36 (finWeyl WeylLabel)
open QAlgorithms.Papers.Independent.Information100.P69 (choi)
open QAlgorithms.Papers.Independent.Information100.P50 (matrixRange)
open QAlgorithms.Papers.Independent.Information100.P08 (IsPure)
attribute [local instance] Classical.propDecidable Classical.decEq
variable {A B C E F R K A2 B2 E2 : Type}
variable [Fintype A] [DecidableEq A] [Nonempty A]
variable [Fintype B] [DecidableEq B] [Nonempty B]
variable [Fintype C] [DecidableEq C] [Nonempty C]
variable [Fintype E] [DecidableEq E] [Nonempty E]
variable [Fintype F] [DecidableEq F] [Nonempty F]
variable [Fintype R] [DecidableEq R] [Nonempty R]
variable [Fintype K] [DecidableEq K]
variable [Fintype A2] [DecidableEq A2] [Nonempty A2]
variable [Fintype B2] [DecidableEq B2] [Nonempty B2]
variable [Fintype E2] [DecidableEq E2] [Nonempty E2]
/-- Appendix A.2, PDF pp.4,26, (72): all joint inputs, different degradable
channels, optimized equality, and every tensor power. Supremum sets are nonempty
and bounded for these nonzero finite systems and isometric channels. -/
theorem coherentInformationAdditive [Nonempty A] [Nonempty B] [Nonempty E]
[Nonempty A2] [Nonempty B2] [Nonempty E2]
(U : Matrix (B × E) A ℂ) (V : Matrix (B2 × E2) A2 ℂ)
(hU : IsDegradable U) (hV : IsDegradable V) :
(∀ rho : Matrix (A × A2) (A × A2) ℂ, IsDensityMatrix rho →
coherentAt (tensorDilation U V) rho ≤
coherentAt U (traceRight rho)+coherentAt V (traceLeft rho)) ∧
maximumCoherentInformation (tensorDilation U V) =
maximumCoherentInformation U+maximumCoherentInformation V ∧
(∀ m : ℕ, maximumCoherentInformation (dilationPower U m) =
(m : ℝ)*maximumCoherentInformation U) := by sorry
end QAlgorithms.Papers.Independent.Gap20.P19
end
Read-back
What the Lean code literally says, in plain math · gpt-6.1-sol
Let be nonempty finite types with decidable equality, and let and be degradable: each satisfies , respectively , and its complementary map is its receiver map followed by some channel, on every input matrix. A channel here is a finite Kraus map with . For any dilation , define and by tracing out its environment and receiver respectively, , and , where is defined by functional calculus. Define the product dilation by , and define . Then every positive semidefinite trace-one matrix on satisfies
and . Also, for every natural number , ; this includes , when the products are empty products equal to , all three registers are singletons, and the asserted value is .
Confirmed by the mission captain (proposal self-audit).
Confirmed by the moderator at approval.