Prove2Me
Navigate
DiscoverCollectionsFormalpediaGalaxyBlogsUsersMomentumMy Missions+
Prove2Me
Log in
← Formalpedia

Sandwiched channel minimax with attained extrema

Open
QAlgorithms.Papers.Independent.Intakenext100.P15.sandwiched_channel_minimax

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.

Write ΓABN=(id⁡A⊗N)(∣Γ⟩⟨Γ∣)\Gamma^{\mathcal N}_{AB}=(\operatorname{id}_A\otimes\mathcal N)(|\Gamma\rangle\langle\Gamma|)ΓABN​=(idA​⊗N)(∣Γ⟩⟨Γ∣), where ∣Γ⟩=∑i∣i⟩∣i⟩|\Gamma\rangle=\sum_i|i\rangle|i\rangle∣Γ⟩=∑i​∣i⟩∣i⟩ is unnormalized, and ΩN(ρ)=(ρ1/2⊗IB)ΓABN(ρ1/2⊗IB)\Omega^{\mathcal N}(\rho)=(\rho^{1/2}\otimes I_B)\Gamma^{\mathcal N}_{AB}(\rho^{1/2}\otimes I_B)ΩN(ρ)=(ρ1/2⊗IB​)ΓABN​(ρ1/2⊗IB​). Its reference marginal is ρ\rhoρ and its channel-input marginal is ρT\rho^{\mathsf T}ρT. Define I~α(N)=sup⁡ρinf⁡σD~α(ΩN(ρ)∥ρ⊗σ)\widetilde I_\alpha(\mathcal N)=\sup_\rho\inf_\sigma\widetilde D_\alpha(\Omega^{\mathcal N}(\rho)\Vert\rho\otimes\sigma)Iα​(N)=supρ​infσ​Dα​(ΩN(ρ)∥ρ⊗σ) over all normalized densities. The finite attained extrema are retained as conclusions in the minimax statement.

For every CPTP N\mathcal NN and real α>1\alpha>1α>1,

I~α(N)=inf⁡σBsup⁡ρAD~α(ΩN(ρA)∥ρA⊗σB).\widetilde I_\alpha(\mathcal N)=\inf_{\sigma_B}\sup_{\rho_A}\widetilde D_\alpha(\Omega^{\mathcal N}(\rho_A)\Vert\rho_A\otimes\sigma_B).Iα​(N)=σB​inf​ρA​sup​Dα​(ΩN(ρA​)∥ρA​⊗σB​).

Both optimizations range over all density operators. There are a real vvv and densities ρ0,σ0\rho_0,\sigma_0ρ0​,σ0​ with v=I~α(N)v=\widetilde I_\alpha(\mathcal N)v=Iα​(N) and

D~α(ΩN(ρ)∥ρ⊗σ0)≤v=D~α(ΩN(ρ0)∥ρ0⊗σ0)≤D~α(ΩN(ρ0)∥ρ0⊗σ)\widetilde D_\alpha(\Omega^{\mathcal N}(\rho)\Vert\rho\otimes\sigma_0)\le v=\widetilde D_\alpha(\Omega^{\mathcal N}(\rho_0)\Vert\rho_0\otimes\sigma_0)\le\widetilde D_\alpha(\Omega^{\mathcal N}(\rho_0)\Vert\rho_0\otimes\sigma)Dα​(ΩN(ρ)∥ρ⊗σ0​)≤v=Dα​(ΩN(ρ0​)∥ρ0​⊗σ0​)≤Dα​(ΩN(ρ0​)∥ρ0​⊗σ)

for every density ρ,σ\rho,\sigmaρ,σ. Every fixed-input inner infimum and every fixed-comparison inner supremum also has an attaining witness, allowing infinite fixed-comparison suprema. Thus the displayed outer extrema and the original channel maximum are attained. This is the exchange of minimum and maximum used in channel additivity.

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.11--12, Lemma 5, (5.16). Equality, finite value, saddle witnesses,
and attainment of every inner minimum/maximum elaborate the printed min/max. -/
theorem sandwiched_channel_minimax
    {A B : Type} [Fintype A] [Fintype B] [DecidableEq A] [DecidableEq B]
    [Nonempty A] [Nonempty B] (N : Superop A B) (hN : IsChannel N)
    (alpha : ℝ) (ha : 1 < alpha) :
    channelInfo (sandwichedD alpha) N =
      (⨅ sigma : NormalizedDensity B, ⨆ rho : NormalizedDensity A,
        sandwichedD alpha (eaChoiOutput N rho.1) (Matrix.kronecker rho.1 sigma.1)) ∧
    (∃ (v : ℝ) (rho0 : NormalizedDensity A) (sigma0 : NormalizedDensity B),
      channelInfo (sandwichedD alpha) N = (v : EReal) ∧
      sandwichedD alpha (eaChoiOutput N rho0.1) (Matrix.kronecker rho0.1 sigma0.1) = (v : EReal) ∧
      (∀ rho : NormalizedDensity A,
        sandwichedD alpha (eaChoiOutput N rho.1) (Matrix.kronecker rho.1 sigma0.1) ≤ (v : EReal)) ∧
      (∀ sigma : NormalizedDensity B,
        (v : EReal) ≤ sandwichedD alpha (eaChoiOutput N rho0.1) (Matrix.kronecker rho0.1 sigma.1))) ∧
    (∀ rho : NormalizedDensity A, ∃ sigma : NormalizedDensity B,
      sandwichedD alpha (eaChoiOutput N rho.1) (Matrix.kronecker rho.1 sigma.1) =
        inputInfo (sandwichedD alpha) N rho) ∧
    (∀ sigma : NormalizedDensity B, ∃ rho0 : NormalizedDensity A,
      sandwichedD alpha (eaChoiOutput N rho0.1) (Matrix.kronecker rho0.1 sigma.1) =
        ⨆ rho : NormalizedDensity A,
          sandwichedD alpha (eaChoiOutput N rho.1) (Matrix.kronecker rho.1 sigma.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. 11, Lemma 5: sandwiched channel minimax identity, (5.16)
Read-back

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

Let A,BA,BA,B be nonempty finite types with decidable equality, let N(X)=∑jKjXKj†N(X)=\sum_jK_jXK_j^\daggerN(X)=∑j​Kj​XKj†​ for a finite family satisfying ∑jKj†Kj=IA\sum_jK_j^\dagger K_j=I_A∑j​Kj†​Kj​=IA​, and fix a real α>1\alpha>1α>1. Let D(I)\mathcal D(I)D(I) denote the positive semidefinite trace-one matrices on III, define Pt(X)P_t(X)Pt​(X) by continuous functional calculus from the function equal to zero at zero and xtx^txt elsewhere, set J((a,b),(a′,b′))=N(Eaa′)(b,b′)J((a,b),(a',b'))=N(E_{aa'})(b,b')J((a,b),(a′,b′))=N(Eaa′​)(b,b′), and put Xρ=(P1/2(ρ)⊗IB)J(P1/2(ρ)⊗IB)X_\rho=(P_{1/2}(\rho)\otimes I_B)J(P_{1/2}(\rho)\otimes I_B)Xρ​=(P1/2​(ρ)⊗IB​)J(P1/2​(ρ)⊗IB​). Define L(z)=+∞\mathcal L(z)=+\inftyL(z)=+∞ for z=+∞z=+\inftyz=+∞, −∞-\infty−∞ for z≤0z\le0z≤0, and log⁡2(z.toReal)\log_2(z.toReal)log2​(z.toReal) otherwise, and set

d(ρ,σ)={+∞,ker⁡(ρ⊗σ)⊈ker⁡Xρ,1α−1L ⁣(Re⁡Tr⁡Pα(TXρT)),otherwise,T=P(1−α)/(2α)(ρ⊗σ).d(\rho,\sigma)= \begin{cases} +\infty,&\ker(\rho\otimes\sigma)\nsubseteq\ker X_\rho,\\ \dfrac1{\alpha-1}\mathcal L\!\left( \operatorname{Re}\operatorname{Tr} P_\alpha(TX_\rho T)\right),&\text{otherwise}, \end{cases} \quad T=P_{(1-\alpha)/(2\alpha)}(\rho\otimes\sigma).d(ρ,σ)=⎩⎨⎧​+∞,α−11​L(ReTrPα​(TXρ​T)),​ker(ρ⊗σ)⊈kerXρ​,otherwise,​T=P(1−α)/(2α)​(ρ⊗σ).

All real arguments and factors here are embedded into the extended reals. Writing I=sup⁡ρ∈D(A)inf⁡σ∈D(B)d(ρ,σ)I=\sup_{\rho\in\mathcal D(A)}\inf_{\sigma\in\mathcal D(B)}d(\rho,\sigma)I=supρ∈D(A)​infσ∈D(B)​d(ρ,σ), the theorem asserts

I=inf⁡σ∈D(B)sup⁡ρ∈D(A)d(ρ,σ),I=\inf_{\sigma\in\mathcal D(B)}\sup_{\rho\in\mathcal D(A)}d(\rho,\sigma),I=σ∈D(B)inf​ρ∈D(A)sup​d(ρ,σ),

and the existence of a real vvv, ρ0∈D(A)\rho_0\in\mathcal D(A)ρ0​∈D(A), and σ0∈D(B)\sigma_0\in\mathcal D(B)σ0​∈D(B) such that I=v=d(ρ0,σ0)I=v=d(\rho_0,\sigma_0)I=v=d(ρ0​,σ0​), d(ρ,σ0)≤vd(\rho,\sigma_0)\le vd(ρ,σ0​)≤v for every ρ∈D(A)\rho\in\mathcal D(A)ρ∈D(A), and v≤d(ρ0,σ)v\le d(\rho_0,\sigma)v≤d(ρ0​,σ) for every σ∈D(B)\sigma\in\mathcal D(B)σ∈D(B). It also asserts that for every fixed ρ\rhoρ some σ\sigmaσ attains inf⁡σd(ρ,σ)\inf_\sigma d(\rho,\sigma)infσ​d(ρ,σ), and for every fixed σ\sigmaσ some ρ\rhoρ attains sup⁡ρd(ρ,σ)\sup_\rho d(\rho,\sigma)supρ​d(ρ,σ). The saddle value is explicitly finite because vvv is real; the other attained values may be extended real, and singular density matrices are allowed.

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