Prove2Me
Navigate
DiscoverCollectionsFormalpediaGalaxyBlogsUsersMomentumMy Missions+
Prove2Me
Log in
← Formalpedia

Corollary 4: optimal channel fidelity is the primal SDP

Proved
QAlgorithms.Papers.Independent.Support50.P11.optimal_fidelity_primal_sdp

by Goku · Oct 23, 2026 · Mathlib 0df444a (Lean v4.33.1)

prove2quantum

Let N:A′→B\mathcal N:A'\to BN:A′→B be a finite CPTP channel and K≥1K\ge1K≥1. Select NS, PPTp, or their intersection using flags s,ps,ps,p with at least one true. Let Ps,p\mathcal P_{s,p}Ps,p​ consist of density matrices ρ\rhoρ and operators Λ\LambdaΛ satisfying 0⪯Λ⪯ρ⊗IB0\preceq\Lambda\preceq\rho\otimes I_B0⪯Λ⪯ρ⊗IB​, together with Tr⁡A′Λ=IB/K2\operatorname{Tr}_{A'}\Lambda=I_B/K^2TrA′​Λ=IB​/K2 when sss is true and −ρ⊗IB/K⪯ΓB(Λ)⪯ρ⊗IB/K-\rho\otimes I_B/K\preceq\Gamma_B(\Lambda)\preceq\rho\otimes I_B/K−ρ⊗IB​/K⪯ΓB​(Λ)⪯ρ⊗IB​/K when ppp is true. For the unnormalized Choi matrix NNN,

FΩs,p(N,K)=max⁡(ρ,Λ)∈Ps,pRe⁡Tr⁡(NTΛ).F^{\Omega_{s,p}}(\mathcal N,K)=\max_{(\rho,\Lambda)\in\mathcal P_{s,p}}\operatorname{Re}\operatorname{Tr}(N^T\Lambda).FΩs,p​(N,K)=(ρ,Λ)∈Ps,p​max​ReTr(NTΛ).

There is a maximizing feasible pair, and an actual code in the selected class attains the same value. The objective is linear and the constraints are affine equalities and PSD inequalities, giving the stated SDP. At K=1K=1K=1 its optimum is one.

Preamble
import Definitions.Def_QAlgorithms_Papers_Independent_Support50_P11
Formal statement
noncomputable section

open scoped Matrix ComplexConjugate ComplexOrder ENNReal

namespace QAlgorithms.Papers.Independent.Support50.P11

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

open QAlgorithms.Papers.Independent.Information100.P21 (pt)

open QAlgorithms.Papers.Independent.Dependency50.P09 (tensorAction)

open QAlgorithms.Papers.Independent.Information100.P76 (swapMatrix)

open QAlgorithms.Papers.Independent.Information100.P69 (replacer)

open QAlgorithms.Papers.Independent.Gap20.P19 (wernerIsometry)

attribute [local instance] Classical.propDecidable

attribute [local instance] Classical.decEq

section Statements

variable {I O : Type} [Fintype I] [Nonempty I] [Fintype O] [Nonempty O]

/-- PDF p.7, Corollary 4: equality with the operational optimum and an attained
primal maximum, for each of NS, PPTp and their intersection, including K=1. -/
theorem optimal_fidelity_primal_sdp
    (N : Superop I O) (hN : IsChannel N) (K : ℕ) (hK : 0 < K)
    (s p : Bool) (hflags : s = true ∨ p = true) :
    ∃ rho : Matrix I I ℂ, ∃ L : Matrix (I × O) (I × O) ℂ,
      Primal K s p rho L ∧ optimalCodeFidelity (flagsClass s p) N K = primalObjective N L ∧
      (∀ rho2 L2, Primal K s p rho2 L2 → primalObjective N L2 ≤ primalObjective N L) ∧
      (∃ Z : Superop (Fin K × O) (I × Fin K),
        IsForwardCode K (flagsClass s p) Z ∧
        channelFidelity K (link K N Z) = primalObjective N L) := by sorry

end Statements

end QAlgorithms.Papers.Independent.Support50.P11

end
Source
On the power of PPT-preserving and non-signalling codes, arXiv:1406.7142v2, PDF p. 7, Corollary 4: optimal channel fidelity is the primal SDP
Read-back

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

For finite index types I,OI,OI,O, with their required finite structures and decidable equality, every channel N:I→ON:I\to ON:I→O, every natural number K>0K>0K>0, and Boolean flags s,ps,ps,p with at least one true, the assertion supplies matrices ρ\rhoρ on III and LLL on I×OI\times OI×O satisfying the conditions below. Here a channel is a function with a finite Kraus representation X↦∑jAjXAj†X\mapsto\sum_jA_jXA_j^\daggerX↦∑j​Aj​XAj†​ for all matrices and ∑jAj†Aj=I\sum_jA_j^\dagger A_j=I∑j​Aj†​Aj​=I; A≤BA\le BA≤B means B−AB-AB−A is positive semidefinite; and Γ\GammaΓ transposes the second register, (ΓX)(a,b),(a′,b′)=X(a,b′),(a′,b)(\Gamma X)_{(a,b),(a',b')}=X_{(a,b'),(a',b)}(ΓX)(a,b),(a′,b′)​=X(a,b′),(a′,b)​. The required matrix conditions are

ρ≥0,Tr⁡ρ=1,0≤L≤ρ⊗IO,\rho\ge0,\qquad \operatorname{Tr}\rho=1,\qquad 0\le L\le\rho\otimes I_O,ρ≥0,Trρ=1,0≤L≤ρ⊗IO​, s=true ⟹ Tr⁡IL=K−2IO,p=true ⟹ −K−1(ρ⊗IO)≤ΓL≤K−1(ρ⊗IO).s=\mathrm{true}\ \Longrightarrow\ \operatorname{Tr}_{I}L=K^{-2}I_O, \qquad p=\mathrm{true}\ \Longrightarrow\ -K^{-1}(\rho\otimes I_O)\le\Gamma L\le K^{-1}(\rho\otimes I_O).s=true ⟹ TrI​L=K−2IO​,p=true ⟹ −K−1(ρ⊗IO​)≤ΓL≤K−1(ρ⊗IO​).

Put J(i,o),(j,o′)=N(Eij)oo′J_{(i,o),(j,o')}=N(E_{ij})_{oo'}J(i,o),(j,o′)​=N(Eij​)oo′​ and v=Re⁡Tr⁡(JTL)v=\operatorname{Re}\operatorname{Tr}(J^{\mathsf T}L)v=ReTr(JTL), using ordinary transpose. A permitted forward code is a channel Z:K×O→I×KZ:K\times O\to I\times KZ:K×O→I×K satisfying, for some channel E:K→IE:K\to IE:K→I, Tr⁡KZ(X)=E(Tr⁡OX)\operatorname{Tr}_{K}Z(X)=E(\operatorname{Tr}_{O}X)TrK​Z(X)=E(TrO​X) for every XXX; when sss is true it additionally satisfies Tr⁡IZ(X)=D(Tr⁡KX)\operatorname{Tr}_{I}Z(X)=D(\operatorname{Tr}_{K}X)TrI​Z(X)=D(TrK​X) for every XXX, for some channel D:O→KD:O\to KD:O→K, and when ppp is true the function Γout∘Z∘Γin\Gamma_{\mathrm{out}}\circ Z\circ\Gamma_{\mathrm{in}}Γout​∘Z∘Γin​ is a channel. For such a ZZZ, define

LN,Z(X)bc=∑a,a′,i,j,β,γXaa′N(Eij)βγZ(E(a,β),(a′,γ))(i,b),(j,c)\mathcal L_{N,Z}(X)_{bc} =\sum_{a,a',i,j,\beta,\gamma} X_{aa'}N(E_{ij})_{\beta\gamma} Z(E_{(a,\beta),(a',\gamma)})_{(i,b),(j,c)}LN,Z​(X)bc​=a,a′,i,j,β,γ∑​Xaa′​N(Eij​)βγ​Z(E(a,β),(a′,γ)​)(i,b),(j,c)​

and its score

FN,Z=Re⁡Tr⁡ ⁣[ΩKΩK†(id⁡K⊗LN,Z)(ΩKΩK†)],ΩK(a,b)=1a=bK.\mathcal F_{N,Z} =\operatorname{Re}\operatorname{Tr}\!\left[ \Omega_K\Omega_K^\dagger (\operatorname{id}_K\otimes\mathcal L_{N,Z})(\Omega_K\Omega_K^\dagger) \right], \qquad \Omega_K(a,b)=\frac{\mathbf1_{a=b}}{\sqrt K}.FN,Z​=ReTr[ΩK​ΩK†​(idK​⊗LN,Z​)(ΩK​ΩK†​)],ΩK​(a,b)=K​1a=b​​.

The theorem states simultaneously that the real supremum of FN,Z\mathcal F_{N,Z}FN,Z​ over all permitted codes is vvv, that every pair ρ2,L2\rho_2,L_2ρ2​,L2​ satisfying the same matrix conditions has Re⁡Tr⁡(JTL2)≤v\operatorname{Re}\operatorname{Tr}(J^{\mathsf T}L_2)\le vReTr(JTL2​)≤v, and that some permitted code ZZZ actually has FN,Z=v\mathcal F_{N,Z}=vFN,Z​=v. It does not require this attaining code’s average input to equal the particular ρ\rhoρ supplied in the conclusion. The statement includes K=1K=1K=1 without a separate branch and excludes only the choice in which both flags are false.

Human review
  • Endorsed by Shuze Chen · Oct 24, 2026

    Confirmed by the moderator at approval.

  • Endorsed by Goku · Oct 24, 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