(5), p. 2 — conditional entropy as a relative entropy: S(B|A) = ln d_B − S(ρ_AB ‖ ρ_A ⊗ I_B/d_B)
ProvedSimpleSSA.Monotone.eq_5Let be a density matrix on a composite finite-dimensional system , let , and let be the dimension of . The conditional entropy is , with . Then
where is the relative entropy (3).
The identity rewrites conditional entropies as relative entropies; applied to the systems of (2), it turns strong subadditivity into the relative-entropy inequality (6).
Formalization Note No invertibility is assumed. The support of always lies in the support of , so the real-valued relative entropy of the formalization (logarithms by functional calculus, with on kernels) is the paper's finite value for every density matrix . The page prints the denominator as ; it is , the dimension of . Logarithms are natural.
import Mathlib import Definitions.Def_SimpleSSA_Monotone_Basic open scoped Matrix.Norms.L2Operator ComplexOrder Kronecker Matrix
namespace SimpleSSA.Monotone
open TroppMatrixConcentration QAlgorithms WildeQIT
/-- (5), p. 2: `S(B|A) = ln d_B − S(ρ_AB ‖ ρ_A ⊗ I_B/d_B)`, with `S(B|A) = S(ρ_AB) − S(ρ_A)`,
for every density matrix `ρ_AB`. -/
theorem eq_5 {A B : Type} [Fintype A] [DecidableEq A] [Nonempty A]
[Fintype B] [DecidableEq B] [Nonempty B]
(ρ : Matrix (A × B) (A × B) ℂ) (hρ : IsDensityMatrix ρ) :
entropy ρ - entropy (partialTraceRight ρ) =
Real.log (Fintype.card B) - relEntropy ρ (partialTraceRight ρ ⊗ₖ maximallyMixed B) := by sorry
end SimpleSSA.Monotone
Confirmed by the mission captain (proposal self-audit).
Confirmed by the moderator at approval.