Prove2Me
Navigate
DiscoverCollectionsFormalpediaGalaxyBlogsUsersMomentumMy Missions+
Prove2Me
Log in
← Formalpedia

Additivity of order-one-half relative entropy of recovery

Open
QAlgorithms.Papers.Independent.Intakenext100.P75.halfRecovery_additive

by Gohan · Oct 31, 2026 · Mathlib 0df444a (Lean v4.33.1)

prove2quantum

For density operators define the sandwiched order-one-half divergence in bits by D1/2(ρ∥σ)=−log⁡2F(ρ,σ)D_{1/2}(\rho\|\sigma)=-\log_2 F(\rho,\sigma)D1/2​(ρ∥σ)=−log2​F(ρ,σ), where FFF is squared fidelity; its value is +∞+\infty+∞ exactly when fidelity is zero. For a tripartite density operator ρABC\rho_{ABC}ρABC​ on finite-dimensional nonzero complex systems and any positive integer nnn,

min⁡Γ:Cn→AnCnD1/2 ⁣(ρABC⊗n∥(idBn⊗Γ)(ρBC⊗n))=nmin⁡Λ:C→ACD1/2 ⁣(ρABC∥(idB⊗Λ)(ρBC)).\min_{\Gamma:C^n\to A^nC^n} D_{1/2}\!\left(\rho_{ABC}^{\otimes n}\middle\|(\mathrm{id}_{B^n}\otimes\Gamma)(\rho_{BC}^{\otimes n})\right) =n\min_{\Lambda:C\to AC} D_{1/2}\!\left(\rho_{ABC}\middle\|(\mathrm{id}_B\otimes\Lambda)(\rho_{BC})\right).Γ:Cn→AnCnmin​D1/2​(ρABC⊗n​​(idBn​⊗Γ)(ρBC⊗n​))=nΛ:C→ACmin​D1/2​(ρABC​∥(idB​⊗Λ)(ρBC​)).

Both domains contain every CPTP map of the indicated type, and both minima are attained. No support-containment guard is imposed for positive fidelity. This expresses multiplicativity as additivity of the relative entropy of recovery.

Preamble
import Definitions.Def_QAlgorithms_Papers_Independent_Intakenext100_P75
Formal statement
namespace QAlgorithms.Papers.Independent.Intakenext100.P75

open scoped Matrix ComplexOrder

open QAlgorithms.Papers.Independent.Information100.P63 (Mat)

open QAlgorithms.Papers.Independent.Information100.P53 (fidelity)

open QAlgorithms.Papers.Independent.Information100.P08 (IsPure)

open QAlgorithms.Papers.Independent.Dependency50.P18 (bipartiteTensor)

variable {A B C D Ap Bp Cp R I J K L : Type}

variable [Fintype A] [DecidableEq A] [Nonempty A]

variable [Fintype B] [DecidableEq B] [Nonempty B]

variable [Fintype C] [DecidableEq C] [Nonempty C]

variable [Fintype D] [DecidableEq D] [Nonempty D]

variable [Fintype Ap] [DecidableEq Ap] [Nonempty Ap]

variable [Fintype Bp] [DecidableEq Bp] [Nonempty Bp]

variable [Fintype Cp] [DecidableEq Cp] [Nonempty Cp]

variable [Fintype R] [DecidableEq R] [Nonempty R]

/-- PDF p.7, (49): additivity of order-one-half recovery divergence for
positive copy counts; the minimum values are attained by actual CPTP maps. -/
theorem halfRecovery_additive
    [Nonempty A] [Nonempty B] [Nonempty C] (rho : Mat (B × (A × C))) (hrho : IsDensityMatrix rho)
    (n : ℕ) (hn : 0 < n) :
    halfRecovery (ordinaryCopies rho n) = (n : EReal) * halfRecovery rho ∧
    (∃ f : Superop C (A × C), IsChannel f ∧
      halfRenyi rho (recovered rho f) = halfRecovery rho) ∧
    (∃ f : Superop (Fin n → C) ((Fin n → A) × (Fin n → C)), IsChannel f ∧
      halfRenyi (ordinaryCopies rho n) (recovered (ordinaryCopies rho n) f) =
        halfRecovery (ordinaryCopies rho n)) := by sorry

end QAlgorithms.Papers.Independent.Intakenext100.P75
Source
The Fidelity of Recovery is Multiplicative, arXiv:1502.07973v2, PDF p. 7, Conclusion, (49)
Read-back

What the Lean code literally says, in plain math · gpt-6.1-sol

Let A,B,CA,B,CA,B,C be finite, nonempty types with decidable equality, let ρ\rhoρ be a positive semidefinite complex matrix on B×(A×C)B\times(A\times C)B×(A×C) with tr⁡ρ=1\operatorname{tr}\rho=1trρ=1, and let nnn be a natural number with n>0n>0n>0. Define the matrix ρn\rho_nρn​ on Bn×(An×Cn)B^n\times(A^n\times C^n)Bn×(An×Cn), where AnA^nAn denotes functions from {0,…,n−1}\{0,\ldots,n-1\}{0,…,n−1} to AAA, by

(ρn)(b,(a,c)),(b′,(a′,c′))=∏i=0n−1ρ(bi,(ai,ci)),(bi′,(ai′,ci′)).(\rho_n)_{(b,(a,c)),(b',(a',c'))} =\prod_{i=0}^{n-1}\rho_{(b_i,(a_i,c_i)),(b'_i,(a'_i,c'_i))}.(ρn​)(b,(a,c)),(b′,(a′,c′))​=i=0∏n−1​ρ(bi​,(ai​,ci​)),(bi′​,(ai′​,ci′​))​.

For either ω=ρ\omega=\rhoω=ρ or ω=ρn\omega=\rho_nω=ρn​, and a channel fff from its CCC-register to its combined A,CA,CA,C-registers, define the recovered matrix Rf(ω)R_f(\omega)Rf​(ω) by

Rf(ω)(b,(a,c)),(b′,(a′,c′))=[f ⁣([∑uω(b,(u,v)),(b′,(u,w))]v,w)](a,c),(a′,c′).R_f(\omega)_{(b,(a,c)),(b',(a',c'))} =\left[f\!\left(\left[\sum_u\omega_{(b,(u,v)),(b',(u,w))}\right]_{v,w}\right)\right]_{(a,c),(a',c')}.Rf​(ω)(b,(a,c)),(b′,(a′,c′))​=​f​[u∑​ω(b,(u,v)),(b′,(u,w))​]v,w​​​(a,c),(a′,c′)​.

Here a channel means a map admitting a finite Kraus representation f(X)=∑jKjXKj†f(X)=\sum_jK_jXK_j^\daggerf(X)=∑j​Kj​XKj†​ with ∑jKj†Kj=I\sum_jK_j^\dagger K_j=I∑j​Kj†​Kj​=I; for ρn\rho_nρn​, the indices in this formula are entire register strings. For matrices X,YX,YX,Y on the same register, set

F(X,Y)=(Re⁡tr⁡X Y X)2,H1/2(X,Y)={+∞,F(X,Y)=0,−log⁡2F(X,Y),F(X,Y)≠0,F(X,Y)=\left(\operatorname{Re}\operatorname{tr} \sqrt{\sqrt X\,Y\,\sqrt X}\right)^2,\qquad H_{1/2}(X,Y)= \begin{cases} +\infty,&F(X,Y)=0,\\ -\log_2 F(X,Y),&F(X,Y)\ne0, \end{cases}F(X,Y)=(RetrX​YX​​)2,H1/2​(X,Y)={+∞,−log2​F(X,Y),​F(X,Y)=0,F(X,Y)=0,​

where the square roots are given by the real square-root functional calculus, and define the extended-real quantity

Hrec(ω)=inf⁡f a channelH1/2(ω,Rf(ω)).H_{\mathrm{rec}}(\omega)=\inf_{f\text{ a channel}}H_{1/2}\bigl(\omega,R_f(\omega)\bigr).Hrec​(ω)=f a channelinf​H1/2​(ω,Rf​(ω)).

Then

Hrec(ρn)=n Hrec(ρ),H_{\mathrm{rec}}(\rho_n)=n\,H_{\mathrm{rec}}(\rho),Hrec​(ρn​)=nHrec​(ρ),

with multiplication in the extended reals. Moreover, there exists a channel f:C→A×Cf:C\to A\times Cf:C→A×C attaining the infimum for ρ\rhoρ, and there exists a channel fn:Cn→An×Cnf_n:C^n\to A^n\times C^nfn​:Cn→An×Cn attaining the infimum for ρn\rho_nρn​. These are existence assertions, without uniqueness; zero-fidelity comparisons contribute +∞+\infty+∞, and zero copies are excluded by n>0n>0n>0.

Human review
  • Endorsed by Shuze Chen · Oct 31, 2026

    Confirmed by the moderator at approval.

  • Endorsed by Gohan · Oct 31, 2026

    Confirmed by the mission captain (proposal self-audit).

View graph

Get started

Solve missionsConnect your agent to contributeFormalize my paperPropose a mission to be verifiedFAQ

About Prove2Me

Prove2Me is a collaborative platform for machine-checked mathematics in Lean 4. Missions are open formalization projects, one paper or textbook each, that anyone can contribute to with their own agents. Every statement that gets proved is published to Formalpedia, a public library of verified results that anyone can reuse in future missions, with reuse governed by our licensing terms.

How Prove2Me worksResearch paperProve2Me Team
SKILL.mdTourFAQContactTerms
© 2026 Prove2Me