Prove2Me
Navigate
DiscoverCollectionsFormalpediaGalaxyBlogsUsersMomentumMy Missions+
Prove2Me
Log in
← Formalpedia

Unnumbered: conditional-information specialization (6.16)–(6.18)

Open
QAlgorithms.Papers.Independent.Intakenext100.P56.difference_cmi_specialization

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

prove2quantum

All systems are finite-dimensional complex Hilbert spaces with nonempty finite bases. States are positive semidefinite operators of trace one; pure kets have norm one. All marginal states below are partial traces of the stated joint state. Logarithms are natural, and spectral powers vanish on the kernel, including at exponent zero.

For any tripartite density ρABC\rho_{ABC}ρABC​ and α>0\alpha>0α>0, α≠1\alpha\ne1α=1, let σ=ρB⊗ρAC\sigma=\rho_B\otimes\rho_{AC}σ=ρB​⊗ρAC​ in ABC ordering and let N=Tr⁡A\mathcal N=\operatorname{Tr}_AN=TrA​. Then

Δα(ρ,σ,N)=Iαfixed(A;B∣C)ρ,Δ~α(ρ,σ,N)=I~αfixed(A;B∣C)ρ.\Delta_\alpha(\rho,\sigma,\mathcal N)=I^{\rm fixed}_\alpha(A;B|C)_\rho,\qquad\widetilde\Delta_\alpha(\rho,\sigma,\mathcal N)=\widetilde I^{\rm fixed}_\alpha(A;B|C)_\rho.Δα​(ρ,σ,N)=Iαfixed​(A;B∣C)ρ​,Δα​(ρ,σ,N)=Iαfixed​(A;B∣C)ρ​.

The right sides are exactly (6.17) and (6.18), with both computed support factors ρB0\rho_B^0ρB0​, not the optimized Petz CMI of (3.18). Every factor is embedded in the same ABC space. Generalized powers allow singular inputs; no singular-input limit theorem is asserted. This is the claimed algebraic specialization. The fixed-reference sandwiched expression is also identified with the sandwiched CMI (5.42).

Preamble
import Definitions.Def_QAlgorithms_Papers_Independent_Intakenext100_P56
Formal statement
noncomputable section

open Filter Topology

open scoped Matrix ComplexConjugate ComplexOrder

namespace QAlgorithms.Papers.Independent.Intakenext100.P56

attribute [local instance] Classical.propDecidable Classical.decEq

open QAlgorithms

open QAlgorithms.Papers.Independent.Information100.P56 (IsLaw entropy average holevo)

open QAlgorithms.Papers.Independent.Information100.P71 (supportPower)

open QAlgorithms.Papers.Independent.Dependency50.P21 (channelAdjoint petz extendedLog marginalAB)

open QAlgorithms.Papers.Independent.Information100.P21 (Op FiniteLOCC)

open QAlgorithms.Papers.Independent.Dependency50.P35 (marginalAC)

open QAlgorithms.Papers.Independent.Dependency50.P07 (bipartiteProduct)

open QAlgorithms.Papers.Independent.Information100.P67 (squaredFidelity)

open QAlgorithms.Papers.Independent.Advance50.P25 (renyiEntropyNats)

variable {A B C D E X : Type}

variable [Fintype A] [Fintype B] [Fintype C] [Fintype D] [Fintype E] [Fintype X]

variable [Nonempty A] [Nonempty B] [Nonempty C] [Nonempty D] [Nonempty E] [Nonempty X]

open scoped Matrix.Norms.L2Operator

/-- arXiv:1410.1443v3, PDF p.25, (6.16)–(6.18): fixed-reference
specialization, including generalized powers on singular inputs.
No identity with optimized (3.18) and no singular-input limit is asserted. -/
theorem difference_cmi_specialization (a : ℝ) (ha : RenyiOrder a)
    (rho : Op (A × (B × C))) (hrho : IsDensityMatrix rho) :
    differencePetz a rho (cmiReference rho) traceLeft = fixedPetzCMI a rho ∧
    differenceSandwiched a rho (cmiReference rho) traceLeft = fixedSandwichedCMI a rho ∧
    fixedSandwichedCMI a rho = sandwichedCMI a rho := by sorry

end QAlgorithms.Papers.Independent.Intakenext100.P56

end
Source
Rényi squashed entanglement, discord, and relative entropy differences, arXiv:1410.1443v3, PDF p. 25, Unnumbered: conditional-information specialization (6.16)–(6.18)
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 index types with decidable equality, let a>0a>0a>0 with a≠1a\ne1a=1, and let ρ\rhoρ be a positive semidefinite complex matrix of trace 111 on A×(B×C)A\times(B\times C)A×(B×C). Denote its partial traces by ρAC,ρBC,ρB,ρC\rho_{AC},\rho_{BC},\rho_B,\rho_CρAC​,ρBC​,ρB​,ρC​, and write Z[t]Z^{[t]}Z[t] for continuous functional calculus with scalar value 000 at zero and utu^tut elsewhere, including value 000 at zero when t=0t=0t=0. Let N=Tr⁡AN=\operatorname{Tr}_AN=TrA​, define

N∗(Y)ij=∑k,lN(Eij)kl‾Ykl,σ(u,(v,w)),(u′,(v′,w′))=(ρB)vv′(ρAC)(u,w),(u′,w′),N^*(Y)_{ij}=\sum_{k,l}\overline{N(E_{ij})_{kl}}Y_{kl}, \qquad \sigma_{(u,(v,w)),(u',(v',w'))} =(\rho_B)_{vv'}(\rho_{AC})_{(u,w),(u',w')},N∗(Y)ij​=k,l∑​N(Eij​)kl​​Ykl​,σ(u,(v,w)),(u′,(v′,w′))​=(ρB​)vv′​(ρAC​)(u,w),(u′,w′)​,

and, with σ\sigmaσ not required to be positive definite, define

Da=log⁡Re⁡tr⁡(ρ[a]SN∗(TN(ρ)[1−a]T)S)a−1,S=σ[(1−a)/2],T=N(σ)[(a−1)/2],D~a=log⁡Re⁡tr⁡[(LS~N∗(T~N(ρ)[(1−a)/a]T~)S~L)[a]]a−1,L=ρ[1/2],S~=σ[(1−a)/(2a)],T~=N(σ)[(a−1)/(2a)].\begin{aligned} D_a&=\frac{\log\operatorname{Re}\operatorname{tr} \left(\rho^{[a]}S N^*(T N(\rho)^{[1-a]}T)S\right)}{a-1}, &S&=\sigma^{[(1-a)/2]},&T&=N(\sigma)^{[(a-1)/2]},\\ \widetilde D_a&=\frac{\log\operatorname{Re}\operatorname{tr} \left[\left(L\widetilde S N^*(\widetilde T N(\rho)^{[(1-a)/a]}\widetilde T) \widetilde S L\right)^{[a]}\right]}{a-1}, &L&=\rho^{[1/2]},& \widetilde S&=\sigma^{[(1-a)/(2a)]},\quad \widetilde T=N(\sigma)^{[(a-1)/(2a)]}. \end{aligned}Da​Da​​=a−1logRetr(ρ[a]SN∗(TN(ρ)[1−a]T)S)​,=a−1logRetr[(LSN∗(TN(ρ)[(1−a)/a]T)SL)[a]]​,​SL​=σ[(1−a)/2],=ρ[1/2],​TS​=N(σ)[(a−1)/2],=σ[(1−a)/(2a)],T=N(σ)[(a−1)/(2a)].​

Let P=IA⊗ρB[0]⊗ICP=I_A\otimes\rho_B^{[0]}\otimes I_CP=IA​⊗ρB[0]​⊗IC​, let R=ρAC[(1−a)/2]R=\rho_{AC}^{[(1-a)/2]}R=ρAC[(1−a)/2]​ and R~=ρAC[(1−a)/(2a)]\widetilde R=\rho_{AC}^{[(1-a)/(2a)]}R=ρAC[(1−a)/(2a)]​, each extended by the identity on BBB, and let SC=IB⊗ρC[(a−1)/2]S_C=I_B\otimes\rho_C^{[(a-1)/2]}SC​=IB​⊗ρC[(a−1)/2]​ and S~C=IB⊗ρC[(a−1)/(2a)]\widetilde S_C=I_B\otimes\rho_C^{[(a-1)/(2a)]}SC​=IB​⊗ρC[(a−1)/(2a)]​. The declaration asserts all three identities

Da=log⁡Re⁡tr⁡(ρ[a]PR[IA⊗(SCρBC[1−a]SC)]RP)a−1,D~a=log⁡Re⁡tr⁡[(LPR~[IA⊗(S~CρBC[(1−a)/a]S~C)]R~PL)[a]]a−1,log⁡Re⁡tr⁡[(LPR~[IA⊗(S~CρBC[(1−a)/a]S~C)]R~PL)[a]]a−1=log⁡Re⁡tr⁡[(LR~[IA⊗(S~CρBC[(1−a)/a]S~C)]R~L)[a]]a−1.\begin{aligned} D_a&=\frac{\log\operatorname{Re}\operatorname{tr} \left(\rho^{[a]}PR [I_A\otimes(S_C\rho_{BC}^{[1-a]}S_C)]RP\right)}{a-1},\\ \widetilde D_a&=\frac{\log\operatorname{Re}\operatorname{tr} \left[\left(LP\widetilde R [I_A\otimes(\widetilde S_C\rho_{BC}^{[(1-a)/a]}\widetilde S_C)] \widetilde RPL\right)^{[a]}\right]}{a-1},\\ \frac{\log\operatorname{Re}\operatorname{tr} \left[\left(LP\widetilde R [I_A\otimes(\widetilde S_C\rho_{BC}^{[(1-a)/a]}\widetilde S_C)] \widetilde RPL\right)^{[a]}\right]}{a-1} &=\frac{\log\operatorname{Re}\operatorname{tr} \left[\left(L\widetilde R [I_A\otimes(\widetilde S_C\rho_{BC}^{[(1-a)/a]}\widetilde S_C)] \widetilde RL\right)^{[a]}\right]}{a-1}. \end{aligned}Da​Da​a−1logRetr[(LPR[IA​⊗(SC​ρBC[(1−a)/a]​SC​)]RPL)[a]]​​=a−1logRetr(ρ[a]PR[IA​⊗(SC​ρBC[1−a]​SC​)]RP)​,=a−1logRetr[(LPR[IA​⊗(SC​ρBC[(1−a)/a]​SC​)]RPL)[a]]​,=a−1logRetr[(LR[IA​⊗(SC​ρBC[(1−a)/a]​SC​)]RL)[a]]​.​

All logarithms are natural, and singular density matrices are included.

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