Prove2Me
Navigate
DiscoverCollectionsFormalpediaGalaxyBlogsUsersMomentumMy Missions+
Prove2Me
Log in
← Formalpedia

§II.C: coherent-information subadditivity and additivity

Open
QAlgorithms.Papers.Independent.Support50.P01.coherentInformation_additivity

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

prove2quantum

Let A=CaA=\mathbb C^aA=Ca, B=CbB=\mathbb C^bB=Cb, and E=CeE=\mathbb C^eE=Ce have positive finite dimensions and fixed orthonormal bases. Fix an isometry U:A→B⊗EU:A\to B\otimes EU:A→B⊗E and define the channel N(X)=Tr⁡E(UXU†)\mathcal N(X)=\operatorname{Tr}_E(UXU^\dagger)N(X)=TrE​(UXU†) and its genuine complement Nc(X)=Tr⁡B(UXU†)\mathcal N^c(X)=\operatorname{Tr}_B(UXU^\dagger)Nc(X)=TrB​(UXU†). Density operators are positive semidefinite with trace one, and all entropies use base-two logarithms with 0log⁡0=00\log 0=00log0=0. Write Ic(N,ρ)=H(N(ρ))−H(Nc(ρ))I_c(\mathcal N,\rho)=H(\mathcal N(\rho))-H(\mathcal N^c(\rho))Ic​(N,ρ)=H(N(ρ))−H(Nc(ρ)), Q(1)(N)=sup⁡ρIc(N,ρ)Q^{(1)}(\mathcal N)=\sup_\rho I_c(\mathcal N,\rho)Q(1)(N)=supρ​Ic​(N,ρ), and Q(N)=sup⁡n≥1Q(1)(N⊗n)/nQ(\mathcal N)=\sup_{n\ge1}Q^{(1)}(\mathcal N^{\otimes n})/nQ(N)=supn≥1​Q(1)(N⊗n)/n, with arbitrary entangled block inputs. Conjugate degradability means one CPTP map DDD satisfies DN(X)=(Nc(X))TD\mathcal N(X)=(\mathcal N^c(X))^TDN(X)=(Nc(X))T for every operator XXX; conjugate antidegradability reverses the receiver and environment in this identity. If N\mathcal NN is conjugate degradable, every two-use density operator ρA1A2\rho_{A_1A_2}ρA1​A2​​, with actual partial-trace marginals ρAi\rho_{A_i}ρAi​​, satisfies

Ic(N⊗2,ρA1A2)≤Ic(N,ρA1)+Ic(N,ρA2),Q(1)(N⊗n)=nQ(1)(N)(n≥1).I_c(\mathcal N^{\otimes2},\rho_{A_1A_2})\le I_c(\mathcal N,\rho_{A_1})+I_c(\mathcal N,\rho_{A_2}),\qquad Q^{(1)}(\mathcal N^{\otimes n})=nQ^{(1)}(\mathcal N)\quad(n\ge1).Ic​(N⊗2,ρA1​A2​​)≤Ic​(N,ρA1​​)+Ic​(N,ρA2​​),Q(1)(N⊗n)=nQ(1)(N)(n≥1).

The first assertion retains entangled inputs, and the second gives optimized additivity.

Preamble
import Definitions.Def_QAlgorithms_Papers_Independent_Support50_P01
Formal statement
noncomputable section

namespace QAlgorithms.Papers.Independent.Support50.P01

open scoped BigOperators Matrix ComplexConjugate ComplexOrder

open Filter

open QAlgorithms.Papers.Independent.Information100.P67
  (receiverChannel complementaryChannel IsDegradable dilationPower maximumCoherentInformation)

open QAlgorithms.Papers.Independent.Information100.P22 (coherentInformation quantumCapacity)

open QAlgorithms.Papers.Independent.Information100.P08 (maximallyMixed)

open QAlgorithms.Papers.Independent.Dependency50.P41 (EntanglementBreaking)

open QAlgorithms.Papers.Independent.Dependency50.P40 (outputFirstChoi)

attribute [local instance] Classical.propDecidable

attribute [local instance] Classical.decEq

variable {a b e c : ℕ}

/-- PDF p.4: subadditivity on arbitrary two-use inputs and optimized n-use additivity. -/
theorem coherentInformation_additivity (ha : 0<a) (hb : 0<b) (he : 0<e)
    (U : Matrix (Fin b × Fin e) (Fin a) ℂ) (hU : ConjugateDegradable U) :
    (∀ rho : Matrix (Fin 2 → Fin a) (Fin 2 → Fin a) ℂ, IsDensityMatrix rho →
      blockCoherentInformation U 2 rho ≤
        coherentInformation (receiverChannel U) (fun x y => ∑ t : Fin a, rho ![x,t] ![y,t]) +
        coherentInformation (receiverChannel U) (fun x y => ∑ t : Fin a, rho ![t,x] ![t,y])) ∧
    ∀ n : ℕ, 0<n → maximumCoherentInformation (dilationPower U n) = (n : ℝ) * maximumCoherentInformation U := by sorry

end QAlgorithms.Papers.Independent.Support50.P01

end
Source
Conjugate Degradability and the Quantum Capacity of Cloning Channels, arXiv:0909.3297v1, PDF p. 4, §II.C: coherent-information subadditivity and additivity
Read-back

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

For all positive natural numbers a,b,ea,b,ea,b,e and every complex matrix U:Ca→Cb⊗CeU:\mathbb C^a\to\mathbb C^b\otimes\mathbb C^eU:Ca→Cb⊗Ce, assume U†U=IaU^\dagger U=I_aU†U=Ia​ and that a finite Kraus channel D:Mat⁡b(C)→Mat⁡e(C)D:\operatorname{Mat}_b(\mathbb C)\to\operatorname{Mat}_e(\mathbb C)D:Matb​(C)→Mate​(C) satisfies D(BU(X))=EU(X)TD(B_U(X))=E_U(X)^{\mathsf T}D(BU​(X))=EU​(X)T for every matrix XXX, where BU(X)=Tr⁡e(UXU†)B_U(X)=\operatorname{Tr}_e(UXU^\dagger)BU​(X)=Tre​(UXU†), EU(X)=Tr⁡b(UXU†)E_U(X)=\operatorname{Tr}_b(UXU^\dagger)EU​(X)=Trb​(UXU†), and “finite Kraus channel” means D(X)=∑jKjXKj†D(X)=\sum_jK_jXK_j^\daggerD(X)=∑j​Kj​XKj†​ with ∑jKj†Kj=Ib\sum_jK_j^\dagger K_j=I_b∑j​Kj†​Kj​=Ib​. Write S(H)=−Re⁡tr⁡[x↦xlog⁡2x](H)S(H)=-\operatorname{Re}\operatorname{tr}[x\mapsto x\log_2x](H)S(H)=−Retr[x↦xlog2​x](H), using continuous functional calculus, and

IΦ(σ)=S(Φ(σ))−S((id⁡⊗Φ)(∣vσ⟩⟨vσ∣)),vσ(i,j)=(σ)j,i.I_\Phi(\sigma)=S(\Phi(\sigma))-S((\operatorname{id}\otimes\Phi)(|v_\sigma\rangle\langle v_\sigma|)),\qquad v_\sigma(i,j)=(\sqrt{\sigma})_{j,i}.IΦ​(σ)=S(Φ(σ))−S((id⊗Φ)(∣vσ​⟩⟨vσ​∣)),vσ​(i,j)=(σ​)j,i​.

Here the identity extension applies Φ\PhiΦ separately to each reference block. For n∈Nn\in\mathbb Nn∈N, let Ub,e;x(n)=∏i=0n−1Ubi,ei;xiU^{(n)}_{\mathbf b,\mathbf e;\mathbf x}=\prod_{i=0}^{n-1}U_{b_i,e_i;x_i}Ub,e;x(n)​=∏i=0n−1​Ubi​,ei​;xi​​, let BU(n)B_{U^{(n)}}BU(n)​ and EU(n)E_{U^{(n)}}EU(n)​ be its two partial-trace maps, and put J(V)=sup⁡σ⪰0, tr⁡σ=1[S(BV(σ))−S(EV(σ))]J(V)=\sup_{\sigma\succeq0,\ \operatorname{tr}\sigma=1}[S(B_V(\sigma))-S(E_V(\sigma))]J(V)=supσ⪰0, trσ=1​[S(BV​(σ))−S(EV​(σ))]. Then every density matrix ρ\rhoρ indexed by two aaa-valued coordinates satisfies

S(BU(2)(ρ))−S(EU(2)(ρ))≤IBU(ρ1)+IBU(ρ2),(ρ1)x,y=∑t=0a−1ρ(x,t),(y,t),(ρ2)x,y=∑t=0a−1ρ(t,x),(t,y).S(B_{U^{(2)}}(\rho))-S(E_{U^{(2)}}(\rho)) \le I_{B_U}(\rho_1)+I_{B_U}(\rho_2), \quad (\rho_1)_{x,y}=\sum_{t=0}^{a-1}\rho_{(x,t),(y,t)},\quad (\rho_2)_{x,y}=\sum_{t=0}^{a-1}\rho_{(t,x),(t,y)}.S(BU(2)​(ρ))−S(EU(2)​(ρ))≤IBU​​(ρ1​)+IBU​​(ρ2​),(ρ1​)x,y​=t=0∑a−1​ρ(x,t),(y,t)​,(ρ2​)x,y​=t=0∑a−1​ρ(t,x),(t,y)​.

Moreover, J(U(n))=nJ(U)J(U^{(n)})=nJ(U)J(U(n))=nJ(U) for every positive natural number nnn; this latter assertion does not quantify over n=0n=0n=0.

Human review
  • Endorsed by Shuze Chen · Oct 23, 2026

    Confirmed by the moderator at approval.

  • Endorsed by Goku · Oct 23, 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