Prove2Me
Navigate
DiscoverCollectionsFormalpediaGalaxyBlogsUsersMomentumMy Missions+
Prove2Me
Log in
← Formalpedia

Consequence 28

Open
QAlgorithms.Papers.Independent.Intakenext100.P56.joint_convexity_remainder_and_equivalence

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

prove2quantum

All registers are finite-dimensional and logarithms are natural. For probability weights pxp_xpx​ and positive definite density matrices ρx,σx\rho_x,\sigma_xρx​,σx​, set ρˉ=∑xpxρx\bar\rho=\sum_xp_x\rho_xρˉ​=∑x​px​ρx​, σˉ=∑xpxσx\bar\sigma=\sum_xp_x\sigma_xσˉ=∑x​px​σx​, and Rx=σx1/2σˉ−1/2ρˉσˉ−1/2σx1/2R_x=\sigma_x^{1/2}\bar\sigma^{-1/2}\bar\rho\bar\sigma^{-1/2}\sigma_x^{1/2}Rx​=σx1/2​σˉ−1/2ρˉ​σˉ−1/2σx1/2​. Conjecture 26 implies the universal joint-convexity remainder inequality

∑xpxD(ρx∥σx)−D(ρˉ∥σˉ)≥−2log⁡∑xpxF(ρx,Rx).\sum_xp_xD(\rho_x\Vert\sigma_x)-D(\bar\rho\Vert\bar\sigma)\ge-2\log\sum_xp_x\sqrt{F(\rho_x,R_x)}.x∑​px​D(ρx​∥σx​)−D(ρˉ​∥σˉ)≥−2logx∑​px​F(ρx​,Rx​)​.

Independently of Conjecture 26, this universal inequality is equivalent to the universal relative-entropy recovery inequality (6.24), on its positive definite state and strict CPTP channel domain. Here FFF is squared fidelity, extended to positive recovered operators without normalizing them; the square root is inside the weighted sum.

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 pp.27–29, Consequence 28, (6.28).
The quantitative bound assumes Conjecture 26. The equivalence between the
UNIVERSAL inequality schemes is unconditional, as its two-direction proof shows. -/
theorem joint_convexity_remainder_and_equivalence :
    (DifferenceParameterConjecture → JointRemainder) ∧
      (JointRemainder ↔ RecoveryRemainder) := by sorry

end QAlgorithms.Papers.Independent.Intakenext100.P56

end
Source
Rényi squashed entanglement, discord, and relative entropy differences, arXiv:1410.1443v3, PDF p. 27, Consequence 28
Read-back

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

The declaration asserts (D⇒J)∧(J⇔R)(\mathcal D\Rightarrow\mathcal J)\land(\mathcal J\Leftrightarrow\mathcal R)(D⇒J)∧(J⇔R), with the three propositions defined as follows. All index types quantified below are finite and nonempty; matrices have complex entries, X†X^\daggerX† denotes conjugate transpose, and a density matrix is a positive semidefinite matrix with trace exactly 111. A channel NNN is a map admitting a finite Kraus representation N(X)=∑jEjXEj†N(X)=\sum_j E_jXE_j^\daggerN(X)=∑j​Ej​XEj†​ with ∑jEj†Ej=I\sum_jE_j^\dagger E_j=I∑j​Ej†​Ej​=I. Write X[t]X^{[t]}X[t] for the continuous functional calculus of the real function that is 000 at 000 and xtx^txt elsewhere, so zero eigenvalues contribute 000 even when t=0t=0t=0. Write log⁡X\log XlogX for the continuous functional calculus of the real logarithm, whose totalized value at 000 is 000, and define

N∗(Y)ij=∑k,lN(∣i⟩⟨j∣)kl‾ Ykl,D0(ρ,σ)=Re⁡tr⁡ ⁣(ρ(log⁡ρ−log⁡σ)),dN(ρ,σ)=D0(ρ,σ)−D0(N(ρ),N(σ)),F(ρ,ω)=[Re⁡tr⁡ρ ω ρ]2,ℓ(x)={−∞,x≤0,log⁡x,x>0.\begin{gathered} N^*(Y)_{ij}=\sum_{k,l}\overline{N(|i\rangle\langle j|)_{kl}}\,Y_{kl},\qquad D_0(\rho,\sigma)=\operatorname{Re}\operatorname{tr}\!\bigl(\rho(\log\rho-\log\sigma)\bigr),\\ d_N(\rho,\sigma)=D_0(\rho,\sigma)-D_0(N(\rho),N(\sigma)),\\ F(\rho,\omega)= \left[\operatorname{Re}\operatorname{tr} \sqrt{\sqrt{\rho}\,\omega\,\sqrt{\rho}}\right]^2,\qquad \ell(x)= \begin{cases} -\infty,&x\le0,\\ \log x,&x>0. \end{cases} \end{gathered}N∗(Y)ij​=k,l∑​N(∣i⟩⟨j∣)kl​​Ykl​,D0​(ρ,σ)=Retr(ρ(logρ−logσ)),dN​(ρ,σ)=D0​(ρ,σ)−D0​(N(ρ),N(σ)),F(ρ,ω)=[Retrρ​ωρ​​]2,ℓ(x)={−∞,logx,​x≤0,x>0.​​

Here logarithms are natural, and expressions involving ℓ\ellℓ and inequalities against it are evaluated in the extended real numbers. The proposition D\mathcal DD says that for every pair of finite nonempty types A,BA,BA,B, every positive definite density matrices ρ,σ\rho,\sigmaρ,σ on AAA, and every channel NNN from matrices on AAA to matrices on BBB satisfying the additional condition that id⁡Fin(n)⊗N\operatorname{id}_{\mathrm{Fin}(n)}\otimes NidFin(n)​⊗N sends every positive definite matrix to a positive definite matrix for every natural number nnn, the following two quantities are nondecreasing in their positive real parameter: for all real 0<a≤b0<a\le b0<a≤b,

Δa(ρ,σ;N)≤Δb(ρ,σ;N),Δ~a(ρ,σ;N)≤Δ~b(ρ,σ;N).\Delta_a(\rho,\sigma;N)\le\Delta_b(\rho,\sigma;N), \qquad \widetilde\Delta_a(\rho,\sigma;N)\le\widetilde\Delta_b(\rho,\sigma;N).Δa​(ρ,σ;N)≤Δb​(ρ,σ;N),Δa​(ρ,σ;N)≤Δb​(ρ,σ;N).

Both quantities equal dN(ρ,σ)d_N(\rho,\sigma)dN​(ρ,σ) at parameter 111; for a≠1a\ne1a=1, setting Sa=σ[(1−a)/2]S_a=\sigma^{[(1-a)/2]}Sa​=σ[(1−a)/2], Ta=N(σ)[(a−1)/2]T_a=N(\sigma)^{[(a-1)/2]}Ta​=N(σ)[(a−1)/2], S~a=σ[(1−a)/(2a)]\widetilde S_a=\sigma^{[(1-a)/(2a)]}Sa​=σ[(1−a)/(2a)], and T~a=N(σ)[(a−1)/(2a)]\widetilde T_a=N(\sigma)^{[(a-1)/(2a)]}Ta​=N(σ)[(a−1)/(2a)], they are

Δa(ρ,σ;N)=log⁡ ⁣(Re⁡tr⁡[ρ[a]SaN∗ ⁣(TaN(ρ)[1−a]Ta)Sa])a−1,Δ~a(ρ,σ;N)=log⁡ ⁣(Re⁡tr⁡[(ρ[1/2]S~aN∗ ⁣(T~aN(ρ)[(1−a)/a]T~a)S~aρ[1/2])[a]])a−1.\begin{aligned} \Delta_a(\rho,\sigma;N) &=\frac{\log\!\left(\operatorname{Re}\operatorname{tr} \left[\rho^{[a]}S_aN^*\!\left(T_aN(\rho)^{[1-a]}T_a\right)S_a\right]\right)}{a-1},\\ \widetilde\Delta_a(\rho,\sigma;N) &=\frac{\log\!\left(\operatorname{Re}\operatorname{tr} \left[\left(\rho^{[1/2]}\widetilde S_a N^*\!\left(\widetilde T_aN(\rho)^{[(1-a)/a]}\widetilde T_a\right) \widetilde S_a\rho^{[1/2]}\right)^{[a]}\right]\right)}{a-1}. \end{aligned}Δa​(ρ,σ;N)Δa​(ρ,σ;N)​=a−1log(Retr[ρ[a]Sa​N∗(Ta​N(ρ)[1−a]Ta​)Sa​])​,=a−1log(Retr[(ρ[1/2]Sa​N∗(Ta​N(ρ)[(1−a)/a]Ta​)Sa​ρ[1/2])[a]])​.​

Thus the parameter comparison includes a=1a=1a=1 or b=1b=1b=1 through the stated branches, and the amplification condition includes n=0n=0n=0. The proposition J\mathcal JJ says that for every finite nonempty matrix index type AAA and finite nonempty label type XXX, every weights px≥0p_x\ge0px​≥0 with ∑xpx=1\sum_xp_x=1∑x​px​=1, and every families of positive definite density matrices ρx,σx\rho_x,\sigma_xρx​,σx​ on AAA, writing ρˉ=∑xpxρx\bar\rho=\sum_xp_x\rho_xρˉ​=∑x​px​ρx​, σˉ=∑xpxσx\bar\sigma=\sum_xp_x\sigma_xσˉ=∑x​px​σx​, and

ρ^x=σx[1/2]σˉ[−1/2]ρˉ σˉ[−1/2]σx[1/2],\widehat\rho_x= \sigma_x^{[1/2]}\bar\sigma^{[-1/2]}\bar\rho\, \bar\sigma^{[-1/2]}\sigma_x^{[1/2]},ρ​x​=σx[1/2]​σˉ[−1/2]ρˉ​σˉ[−1/2]σx[1/2]​,

one has

−2ℓ ⁣(∑xpxF(ρx,ρ^x))≤∑xpxD0(ρx,σx)−D0(ρˉ,σˉ).-2\ell\!\left(\sum_xp_x\sqrt{F(\rho_x,\widehat\rho_x)}\right) \le \sum_xp_xD_0(\rho_x,\sigma_x)-D_0(\bar\rho,\bar\sigma).−2ℓ(x∑​px​F(ρx​,ρ​x​)​)≤x∑​px​D0​(ρx​,σx​)−D0​(ρˉ​,σˉ).

Zero weights are allowed, but the positive definiteness and density requirements still apply to every label. Finally, R\mathcal RR says that for every finite nonempty A,BA,BA,B, positive definite density matrices ρ,σ\rho,\sigmaρ,σ on AAA, and channel NNN satisfying the same positive definiteness condition for every amplification, the map

Pσ,N(Y)=σ N∗ ⁣((N(σ))−1Y(N(σ))−1)σ,\mathcal P_{\sigma,N}(Y)= \sqrt{\sigma}\, N^*\!\left((\sqrt{N(\sigma)})^{-1}Y(\sqrt{N(\sigma)})^{-1}\right) \sqrt{\sigma},Pσ,N​(Y)=σ​N∗((N(σ)​)−1Y(N(σ)​)−1)σ​,

where inverse square roots use functional calculus with value 000 at zero, satisfies

−ℓ ⁣(F(ρ,Pσ,N(N(ρ))))≤dN(ρ,σ).-\ell\!\left(F\bigl(\rho,\mathcal P_{\sigma,N}(N(\rho))\bigr)\right) \le d_N(\rho,\sigma).−ℓ(F(ρ,Pσ,N​(N(ρ))))≤dN​(ρ,σ).

In these remainder inequalities a nonpositive argument of ℓ\ellℓ makes the left side +∞+\infty+∞. The declaration therefore gives the implication from the specified parameter monotonicity proposition to the specified joint remainder proposition, together with both directions of equivalence between that joint remainder proposition and the specified recovery remainder proposition.

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