Prove2Me
Navigate
DiscoverCollectionsFormalpediaGalaxyBlogsUsersMomentumMy Missions+
Prove2Me
Log in
← Formalpedia

Right derivative limit of the Petz log-trace function

Open
QAlgorithms.Papers.Independent.Intakenext100.P15.logtrace_derivative_limit

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

prove2quantum

All Hilbert spaces are finite-dimensional and nonzero. A density operator is positive semidefinite with trace one; singular states are included. All logarithms are base two. For real α>1\alpha>1α>1, D~α\widetilde D_\alphaDα​ denotes the sandwiched Rényi divergence and DαD_\alphaDα​ the traditional Petz divergence, with value +∞+\infty+∞ on support failure and negative spectral powers taken on the support.

For a bipartite density ρAB\rho_{AB}ρAB​, let ρA=Tr⁡BρAB\rho_A=\operatorname{Tr}_B\rho_{AB}ρA​=TrB​ρAB​ and put Kα=Tr⁡A[((ρA(1−α)/2)⊗IB)ρABα((ρA(1−α)/2)⊗IB)]K_\alpha=\operatorname{Tr}_A[((\rho_A^{(1-\alpha)/2})\otimes I_B)\rho_{AB}^\alpha((\rho_A^{(1-\alpha)/2})\otimes I_B)]Kα​=TrA​[((ρA(1−α)/2​)⊗IB​)ρABα​((ρA(1−α)/2​)⊗IB​)], Zα=Tr⁡Kα1/αZ_\alpha=\operatorname{Tr}K_\alpha^{1/\alpha}Zα​=TrKα1/α​, and σB∗=Kα1/α/Zα\sigma_B^*=K_\alpha^{1/\alpha}/Z_\alphaσB∗​=Kα1/α​/Zα​. Partial-trace cyclicity gives Kα=Tr⁡A[(ρA1−α⊗IB)ρABα]K_\alpha=\operatorname{Tr}_A[(\rho_A^{1-\alpha}\otimes I_B)\rho_{AB}^\alpha]Kα​=TrA​[(ρA1−α​⊗IB​)ρABα​].

Put f(α)=log⁡2Zαf(\alpha)=\log_2 Z_\alphaf(α)=log2​Zα​. This function is differentiable at every α>1\alpha>1α>1, and

lim⁡α↓1f′(α)=S(ρA)+S(ρB)−S(ρAB)=I(A;B)ρ,\lim_{\alpha\downarrow1} f'(\alpha)=S(\rho_A)+S(\rho_B)-S(\rho_{AB})=I(A;B)_\rho,α↓1lim​f′(α)=S(ρA​)+S(ρB​)−S(ρAB​)=I(A;B)ρ​,

where ρB=Tr⁡AρAB\rho_B=\operatorname{Tr}_A\rho_{AB}ρB​=TrA​ρAB​ and S(τ)=−Tr⁡(τlog⁡2τ)S(\tau)=-\operatorname{Tr}(\tau\log_2\tau)S(τ)=−Tr(τlog2​τ). The limit concerns the derivative, and approaches one from above. This identifies the first-order behavior used for positive converse payoffs.

Preamble
import Definitions.Def_QAlgorithms_Papers_Independent_Intakenext100_P15
Formal statement
noncomputable section

open scoped Matrix ComplexOrder ComplexConjugate Topology

open Filter

namespace QAlgorithms.Papers.Independent.Intakenext100.P15

attribute [local instance] Classical.propDecidable

attribute [local instance] Classical.decEq

-- These are pinned accepted definitions, not other chunks from the current run.

open QAlgorithms.Papers.Independent.Bridge50.P43 (NormalizedDensity payoff)

open QAlgorithms.Papers.Independent.Dependency50.P09 (Divergence logExt sandwichedD petzD)

open QAlgorithms.Papers.Independent.Information100.P34 (choi)

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

open QAlgorithms.Papers.Independent.Information100.P71
  (supportPower SupportIncluded stateRelative channelPower diamond)

/-- PDF pp.15--16, Lemma 9, (6.25). The derivative approaches order one from
above; both the log-trace function and mutual information use base two. -/
theorem logtrace_derivative_limit
    {A B : Type} [Fintype A] [Fintype B] [DecidableEq A] [DecidableEq B]
    [Nonempty A] [Nonempty B] (rho : NormalizedDensity (A × B)) :
    (∀ alpha : ℝ, 1 < alpha → DifferentiableAt ℝ (logTrace rho.1) alpha) ∧
    Tendsto (deriv (logTrace rho.1)) (𝓝[>] (1 : ℝ)) (𝓝 (ordinaryStateInfo rho.1)) := by sorry

end QAlgorithms.Papers.Independent.Intakenext100.P15

end
Source
Multiplicativity of completely bounded ppp-norms implies a strong converse for entanglement-assisted capacity, arXiv:1310.7028v3, PDF p. 15, Lemma 9: derivative limit at order one, (6.25)
Read-back

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

Let A,BA,BA,B be finite nonempty index sets with decidable equality, and let ρ\rhoρ be a positive semidefinite trace-one matrix on A×BA\times BA×B. Write ρA=Tr⁡Bρ\rho_A=\operatorname{Tr}_B\rhoρA​=TrB​ρ and ρB=Tr⁡Aρ\rho_B=\operatorname{Tr}_A\rhoρB​=TrA​ρ, where each partial trace sums over the diagonal index of the discarded factor. For a positive semidefinite matrix XXX, let X[t]X^{[t]}X[t] be its spectral functional calculus with scalar value 000 at 000 and xtx^txt at every nonzero eigenvalue. Define the real-valued function

f(α)=log⁡2 ⁣(Re⁡Tr⁡{Tr⁡A ⁣[(ρA[(1−α)/2]⊗IB)ρ[α](ρA[(1−α)/2]⊗IB)]}[1/α]).f(\alpha)=\log_2\!\left( \operatorname{Re}\operatorname{Tr} \left\{ \operatorname{Tr}_A\!\left[ (\rho_A^{[(1-\alpha)/2]}\otimes I_B)\rho^{[\alpha]} (\rho_A^{[(1-\alpha)/2]}\otimes I_B) \right]\right\}^{[1/\alpha]} \right).f(α)=log2​(ReTr{TrA​[(ρA[(1−α)/2]​⊗IB​)ρ[α](ρA[(1−α)/2]​⊗IB​)]}[1/α]).

Here the outer matrix power is applied to the matrix inside braces, and the real logarithm is total, with log⁡20=0\log_2 0=0log2​0=0. Define S(X)=−Re⁡Tr⁡g(X)S(X)=-\operatorname{Re}\operatorname{Tr}g(X)S(X)=−ReTrg(X), where g(x)=xlog⁡2xg(x)=x\log_2xg(x)=xlog2​x is applied by functional calculus and g(0)=0g(0)=0g(0)=0. The assertion is that fff is differentiable as a real function at every α>1\alpha>1α>1, and that

lim⁡α→1+f′(α)=S(ρA)+S(ρB)−S(ρ).\lim_{\alpha\to1^+}f'(\alpha) = S(\rho_A)+S(\rho_B)-S(\rho).α→1+lim​f′(α)=S(ρA​)+S(ρB​)−S(ρ).

The derivative limit is from above; no invertibility assumption is imposed on ρ\rhoρ or its marginals.

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