Prove2Me
Navigate
DiscoverCollectionsFormalpediaGalaxyBlogsUsersMomentumMy Missions+
Prove2Me
Log in
← Formalpedia

§5.4 (62)–(63): superadditivity and optimal-average factorization

Open
QAlgorithms.Papers.Independent.Gap48.P18.joint_set_capacity_superadditivity

by Goku · Oct 22, 2026 · Mathlib 0df444a (Lean v4.33.1)

prove2quantum

Let A,BA,BA,B be nonempty marginal state sets, and let A⊗BA\otimes BA⊗B be all joint states with marginals in those sets. Including infinite values,

Cˉ(A⊗B)≥Cˉ(A)+Cˉ(B).\bar C(A\otimes B)\ge\bar C(A)+\bar C(B).Cˉ(A⊗B)≥Cˉ(A)+Cˉ(B).

When both marginal capacities are finite and equality holds, Ω(A⊗B)=Ω(A)⊗Ω(B)\Omega(A\otimes B)=\Omega(A)\otimes\Omega(B)Ω(A⊗B)=Ω(A)⊗Ω(B). Strict inequality is possible: for identical infinite-entropy singleton marginals the joint capacity is infinite whereas both singleton capacities are zero. Conversely, even finite joint capacity and a product optimal average do not imply equality: identical uniform rank-two singleton marginals give such a counterexample. This concerns all joint states, rather than a product-state-only set or a tensor channel's output set.

Preamble
import Definitions.Def_QAlgorithms_Papers_Independent_Gap48_P18
Formal statement
noncomputable section

open MeasureTheory Filter

open scoped ENNReal NNReal Topology Matrix.Norms.L2Operator

open QAlgorithms.Papers.Independent.Dependency50.P33
  (State PureState traceSize traceClosure IsTraceClassChannel productCoeff)

open QAlgorithms.Papers.Independent.Frontier18.P07
  (GeneralizedEnsemble HasFiniteSupport HasAverage averageCoefficients
   extendedEntropyNats extendedRelativeEntropyNats
   finiteEnsembleCapacity pureStates)

open QAlgorithms.Papers.Independent.Frontier18.P01 (traceFirst traceSecond)

namespace QAlgorithms.Papers.Independent.Gap48.P18

attribute [local instance] Classical.propDecidable Classical.decEq

/-- §5.4 (62)–(63): superadditivity and optimal-average factorization, arXiv:quant-ph/0510073v2, PDF pp.63.
The infinite-entropy isomorphic-singleton example is represented with identical coordinate copies. A finite rank-two product-center counterexample witnesses failure of the converse. -/
theorem joint_set_capacity_superadditivity {I J : Type} [Countable I] [Countable J]
    [Nonempty I] [Nonempty J] (A : Set (State I)) (B : Set (State J))
    (hA : A.Nonempty) (hB : B.Nonempty) :
    setChiCapacity A + setChiCapacity B ≤ setChiCapacity (jointMarginalConstraintSet A B) ∧
      (∀ hAB : (jointMarginalConstraintSet A B).Nonempty, setChiCapacity A < ⊤ → setChiCapacity B < ⊤ →
        setChiCapacity (jointMarginalConstraintSet A B) = setChiCapacity A + setChiCapacity B →
        (optimalAverage (jointMarginalConstraintSet A B) hAB).coefficients =
          productCoeff (optimalAverage A hA).coefficients (optimalAverage B hB).coefficients) ∧
      (∀ rho : State I, stateEntropyNats rho = ⊤ → setChiCapacity (couplingSet rho rho) = ⊤ ∧
        setChiCapacity ({rho} : Set (State I)) = 0) ∧
      (∃ (rho : State (Fin 2)) (hC : (couplingSet rho rho).Nonempty),
        rho.coefficients = (1 / 2 : ℂ) • (1 : Matrix (Fin 2) (Fin 2) ℂ) ∧
        setChiCapacity ({rho} : Set (State (Fin 2))) = 0 ∧
        0 < setChiCapacity (couplingSet rho rho) ∧ setChiCapacity (couplingSet rho rho) < ⊤ ∧
        (optimalAverage (couplingSet rho rho) hC).coefficients =
          productCoeff rho.coefficients rho.coefficients) := by sorry

end QAlgorithms.Papers.Independent.Gap48.P18

end
Source
Entropic characteristics of subsets of states, arXiv:quant-ph/0510073v2, PDF p. 63, §5.4 (62)–(63): superadditivity and optimal-average factorization
Read-back

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

Let I,JI,JI,J be countable nonempty label sets, and let A,BA,BA,B be nonempty sets of states on I,JI,JI,J, respectively, with specified nonemptiness witnesses. For any label set EEE, a state XXX is a complex matrix admitting an expansion Xij=∑aua(i)va(j)‾X_{ij}=\sum_a u_a(i)\overline{v_a(j)}Xij​=∑a​ua​(i)va​(j)​, where ua,va∈ℓ2(E;C)u_a,v_a\in\ell^2(E;\mathbb C)ua​,va​∈ℓ2(E;C) and ∑a∥ua∥2∥va∥2<∞\sum_a\|u_a\|_2\|v_a\|_2<\infty∑a​∥ua​∥2​∥va​∥2​<∞, with real nonnegative finite quadratic forms and summable real diagonal of sum 111. Let τ(X)\tau(X)τ(X) be the infimum of those sums over such expansions, with empty infimum +∞+\infty+∞, and give states the topology generated by {Y:τ(Y−X)<ε}\{Y:\tau(Y-X)<\varepsilon\}{Y:τ(Y−X)<ε}, for states XXX and ε>0\varepsilon>0ε>0, and its Borel measurable structure. Define

D(X∥Y)=sup⁡s finite{[ℜtr⁡(Xslog⁡Xs−Xslog⁡Ys+Ys−Xs)]+,ker⁡Ys⊆ker⁡Xs,+∞,otherwise,H(X)=sup⁡s finite[(ℜtr⁡Xs)log⁡(ℜtr⁡Xs)−ℜtr⁡(Xslog⁡Xs)]+,D(X\|Y)=\sup_{s\text{ finite}} \begin{cases} \bigl[\Re\operatorname{tr}(X_s\log X_s-X_s\log Y_s+Y_s-X_s)\bigr]_+, &\ker Y_s\subseteq\ker X_s,\\ +\infty,&\text{otherwise}, \end{cases} \quad H(X)=\sup_{s\text{ finite}} \bigl[(\Re\operatorname{tr}X_s)\log(\Re\operatorname{tr}X_s) -\Re\operatorname{tr}(X_s\log X_s)\bigr]_+,D(X∥Y)=s finitesup​{[ℜtr(Xs​logXs​−Xs​logYs​+Ys​−Xs​)]+​,+∞,​kerYs​⊆kerXs​,otherwise,​H(X)=s finitesup​[(ℜtrXs​)log(ℜtrXs​)−ℜtr(Xs​logXs​)]+​,

where Xs,YsX_s,Y_sXs​,Ys​ are finite principal compressions, [r]+=max⁡(r,0)[r]_+=\max(r,0)[r]+​=max(r,0), and matrix logarithms use continuous functional calculus with the total natural logarithm. For a measure μ\muμ, let (Xˉμ)ij=∫Xij dμ(X)(\bar X_\mu)_{ij}=\int X_{ij}\,d\mu(X)(Xˉμ​)ij​=∫Xij​dμ(X) and χ(μ)=∫D(X∥Xˉμ) dμ(X)\chi(\mu)=\int D(X\|\bar X_\mu)\,d\mu(X)χ(μ)=∫D(X∥Xˉμ​)dμ(X). Let C(S)C(S)C(S) be the supremum of χ(μ)\chi(\mu)χ(μ) over Borel probability measures with finite topological support contained in SSS, with empty supremum 000. For a nonempty SSS and a specified nonemptiness witness, let OSO_SOS​ select a state ω\omegaω with D(X∥ω)≤C(S)D(X\|\omega)\le C(S)D(X∥ω)≤C(S) for all X∈SX\in SX∈S, if one exists, and otherwise select a member of SSS using that witness. Let M(A,B)M(A,B)M(A,B) consist of states Ω\OmegaΩ on I×JI\times JI×J for which some ρ∈A\rho\in Aρ∈A and σ∈B\sigma\in Bσ∈B satisfy

∑r∈J′Ω(i,r),(j,r)=ρij,∑r∈I′Ω(r,i),(r,j)=σij\sum_{r\in J}'\Omega_{(i,r),(j,r)}=\rho_{ij}, \qquad \sum_{r\in I}'\Omega_{(r,i),(r,j)}=\sigma_{ij}r∈J∑′​Ω(i,r),(j,r)​=ρij​,r∈I∑′​Ω(r,i),(r,j)​=σij​

for all relevant indices, and put K(ρ,σ)=M({ρ},{σ})K(\rho,\sigma)=M(\{\rho\},\{\sigma\})K(ρ,σ)=M({ρ},{σ}). Four assertions hold. First,

C(A)+C(B)≤C(M(A,B)).C(A)+C(B)\le C(M(A,B)).C(A)+C(B)≤C(M(A,B)).

Second, for every nonemptiness witness for M(A,B)M(A,B)M(A,B), if C(A)C(A)C(A) and C(B)C(B)C(B) are finite and C(M(A,B))=C(A)+C(B)C(M(A,B))=C(A)+C(B)C(M(A,B))=C(A)+C(B), then the selected state has coefficients

(OM(A,B))(i,k),(j,l)=(OA)ij(OB)kl.(O_{M(A,B)})_{(i,k),(j,l)}=(O_A)_{ij}(O_B)_{kl}.(OM(A,B)​)(i,k),(j,l)​=(OA​)ij​(OB​)kl​.

Third, for every state ρ\rhoρ on III, H(ρ)=+∞H(\rho)=+\inftyH(ρ)=+∞ implies C(K(ρ,ρ))=+∞C(K(\rho,\rho))=+\inftyC(K(ρ,ρ))=+∞ and C({ρ})=0C(\{\rho\})=0C({ρ})=0. Fourth, there exist a state ρ\rhoρ on the two-element set {0,1}\{0,1\}{0,1} and a nonemptiness witness for K(ρ,ρ)K(\rho,\rho)K(ρ,ρ) such that

ρ=121,C({ρ})=0,0<C(K(ρ,ρ))<+∞,(OK(ρ,ρ))(i,k),(j,l)=ρijρkl.\rho=\tfrac12\mathbf1,\qquad C(\{\rho\})=0,\qquad 0<C(K(\rho,\rho))<+\infty,\qquad (O_{K(\rho,\rho)})_{(i,k),(j,l)}=\rho_{ij}\rho_{kl}.ρ=21​1,C({ρ})=0,0<C(K(ρ,ρ))<+∞,(OK(ρ,ρ)​)(i,k),(j,l)​=ρij​ρkl​.

The first inequality permits infinite capacities; finiteness is imposed only in the indicated later clauses.

Human review
  • Endorsed by Shuze Chen · Oct 22, 2026

    Confirmed by the moderator at approval.

  • Endorsed by Goku · Oct 22, 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