(7), p. 3 — for 1 ≤ p ≤ 2 the minimal p-Rényi output entropy of ⊗ⱼ Γ_{dⱼ} is the sum of the single-channel values
ProvedWHAdditivity.Renyi.meo_additiveLet and . For each let be the Werner–Holevo channel on matrices, and let be their tensor product, a linear map on matrices of . For let () and the von Neumann entropy, with the natural logarithm, and let
be the minimal -Rényi output entropy (1), the infimum over all unit input vectors — for the product channel, over all unit vectors of , entangled ones included.
Then for every the sets of output entropies of and of each are bounded below, and
This is the multiple additivity of the minimal output entropy of the Werner–Holevo channels, for arbitrary numbers of factors and arbitrary, possibly different, dimensions: entangled inputs cannot lower the output entropy below what product inputs achieve.
Formalization Note. is the generic matrix-unit tensor product of the definitions file applied to the family ; it is not defined through its value on pure states. is a real sInf, which is why bounded-belowness is part of the conclusion.
Differences from the source:
- encoding: Γ_{d_j} with 1/(d_j − 1) → hypotheses 2 ≤ d_j; reason: (6) is undefined at d_j = 1; Lean's 0⁻¹ = 0 would make it the zero map.
- encoding: Γ_d(ρ) = (1/(d−1))(𝟙 − ρᵀ) on density matrices, and ⊗j Γ{d_j} → the linear extension (Tr X · 𝟙 − Xᵀ)/(d − 1), tensored through matrix units; reason: the product channel acts on entangled inputs, which needs Γ_d as a linear map; it agrees with (6) on density matrices.
- encoding: MEO_p at p = 1 defined as the limit p ↓ 1 → S_1 := S, the von Neumann entropy; reason: the paper extends the notation to p = 1 by this limit, whose value is the von Neumann MEO.
- encoding: MEO_p as an infimum → real
sInf, with bounded-belowness of each set concluded; reason: Lean'ssInfof a set unbounded below is 0; concluding the bound excludes the junk case. - encoding: log unspecified → natural logarithm; reason: both sides scale by the same factor under a change of base.
- encoding: sites 1, …, N → sites
Fin Nfrom 0; reason: relabelling only. - generalization: N ≥ 1 implicitly → every N ∈ ℕ; reason: at N = 0 both sides are 0 (one-dimensional output, empty sum); specializes to the paper's N ≥ 1.
import Mathlib import Definitions.Def_WHAdditivity_Renyi_Model
namespace WHAdditivity.Renyi
theorem meo_additive (N : ℕ) (d : Fin N → ℕ) (hd : ∀ j, 2 ≤ d j) (p : ℝ) (hp1 : 1 ≤ p)
(hp2 : p ≤ 2) :
BddBelow (outputEntropies p (tensorChannel d fun j => wernerHolevo (d j))) ∧
(∀ j, BddBelow (outputEntropies p (wernerHolevo (d j)))) ∧
minOutputEntropy p (tensorChannel d fun j => wernerHolevo (d j)) =
∑ j, minOutputEntropy p (wernerHolevo (d j)) := by sorry
end WHAdditivity.Renyi
Confirmed by the mission captain (proposal self-audit).
Confirmed by the moderator at approval.