Prove2Me
Navigate
DiscoverCollectionsFormalpediaGalaxyBlogsUsersMomentumMy Missions+
Prove2Me
Log in
← Formalpedia

Display (1) (p. 3) — I(σ̃) = I({p_ij}) + Σ p_i^A S(τ^i) + Σ p_j^B S(τ^j) − Σ p_ij S(ρ^ij) ≥ I({p_ij}) + Σ p_ij I(ρ^ij)

Proved
NoLocalBroadcast.Broadcast.display_1

by mikedeng1 · Oct 24, 2026 · Mathlib 0df444a (Lean v4.33.1)

concavitymutual-informationp2o-batch-pfp2ap2o-gran-per-chapterp2o-plan-paperp2o-strengthenedp2o-v1von-neumann-entropy

Let III, JJJ be finite sets, {pij}\{p_{ij}\}{pij​} a probability distribution on I×JI\times JI×J, and ρij\rho^{ij}ρij states of a finite-dimensional bipartite system A′B′A'B'A′B′. Let

σ~=∑ijpij ∣i⟩⟨i∣A⊗∣j⟩⟨j∣B⊗ρA′B′ij,\tilde\sigma=\sum_{ij}p_{ij}\,|i\rangle\langle i|_A\otimes|j\rangle\langle j|_B\otimes\rho^{ij}_{A'B'},σ~=ij∑​pij​∣i⟩⟨i∣A​⊗∣j⟩⟨j∣B​⊗ρA′B′ij​,

regarded as a state across the cut AA′:BB′AA':BB'AA′:BB′. Write piA=∑jpijp_i^A=\sum_jp_{ij}piA​=∑j​pij​, pjB=∑ipijp_j^B=\sum_ip_{ij}pjB​=∑i​pij​, τA′i=∑jpijpiAρA′ij\tau^i_{A'}=\sum_j\frac{p_{ij}}{p_i^A}\rho^{ij}_{A'}τA′i​=∑j​piA​pij​​ρA′ij​ and τB′j=∑ipijpjBρB′ij\tau^j_{B'}=\sum_i\frac{p_{ij}}{p_j^B}\rho^{ij}_{B'}τB′j​=∑i​pjB​pij​​ρB′ij​, and let I({pij})I(\{p_{ij}\})I({pij​}) be the classical mutual information of {pij}\{p_{ij}\}{pij​}. Then

I(σ~)=I({pij})+∑ipiAS(τA′i)+∑jpjBS(τB′j)−∑ijpijS(ρA′B′ij) ≥ I({pij})+∑ijpijI(ρA′B′ij).I(\tilde\sigma)=I(\{p_{ij}\})+\sum_ip_i^AS(\tau^i_{A'})+\sum_jp_j^BS(\tau^j_{B'})-\sum_{ij}p_{ij}S(\rho^{ij}_{A'B'})\ \ge\ I(\{p_{ij}\})+\sum_{ij}p_{ij}I(\rho^{ij}_{A'B'}).I(σ~)=I({pij​})+i∑​piA​S(τA′i​)+j∑​pjB​S(τB′j​)−ij∑​pij​S(ρA′B′ij​) ≥ I({pij​})+ij∑​pij​I(ρA′B′ij​).

In the proof of Theorem 3, σ~\tilde\sigmaσ~ is obtained by measuring AAA and BBB of a broadcast state with measurements optimal for ICC(ρ)I_{CC}(\rho)ICC​(ρ), so that I({pij})=ICC(ρ)I(\{p_{ij}\})=I_{CC}(\rho)I({pij​})=ICC​(ρ) and the lower bound is the printed ICC(ρ)+∑ijpijI(ρij)I_{CC}(\rho)+\sum_{ij}p_{ij}I(\rho^{ij})ICC​(ρ)+∑ij​pij​I(ρij).

Formalization Note σ~\tilde\sigmaσ~ is flaggedState p ρij, indexed ((i,x′),(j,y′))((i,x'),(j,y'))((i,x′),(j,y′)). I({pij})I(\{p_{ij}\})I({pij​}) is the quantum mutual information of the diagonal matrix of ppp on I×JI\times JI×J, which equals the classical mutual information. When piA=0p_i^A=0piA​=0 the division gives τA′i=0\tau^i_{A'}=0τA′i​=0, whose entropy is 000, and the term is multiplied by piA=0p_i^A=0piA​=0 anyway; similarly for BBB. The ρij\rho^{ij}ρij are states for every (i,j)(i,j)(i,j); those with pij=0p_{ij}=0pij​=0 do not enter σ~\tilde\sigmaσ~.

Differences from the source:

  • generalization: the specific σ̃ = (M_A ⊗ N_B)[σ] of the proof of Theorem 3 → every σ̃ = Σ_ij p_ij |i⟩⟨i| ⊗ |j⟩⟨j| ⊗ ρ^ij for finite registers, a probability distribution p and states ρ^ij; reason: the proof's σ̃ is of this form, so the statement specializes to it
  • generalization: lower bound I_CC(ρ) + Σ_ij p_ij I(ρ^ij) → lower bound I({p_ij}) + Σ_ij p_ij I(ρ^ij); reason: the paper replaces I({p_ij}) by I_CC(ρ) using I({p_ij}) = I_CC(ρ) for its optimal measurements; in that context the two bounds coincide, and the Lean bound holds for every σ̃
  • encoding: classical mutual information I({p_ij}) = H(p^A) + H(p^B) − H(p) → mutualInfo of Matrix.diagonal p on ι × κ; reason: the von Neumann entropies of diagonal matrices are the Shannon entropies of their diagonals; same number
  • encoding: ρ^ij defined where p_ij > 0; τ^i with p_i^A in the denominator → ρ^ij a state for every (i, j); real division with x / 0 = 0; reason: terms with p_ij = 0 or p_i^A = 0 carry weight 0 on both sides; same values
  • encoding: log of unspecified base in S(ρ) = −Tr ρ log ρ → QAlgorithms.vnEntropy, base 2; reason: every statement here is invariant under the base of the logarithm
  • encoding: σ̃ = Σ_ij p_ij |i_A j_B⟩⟨i_A j_B| ⊗ ρ^ij_{A′B′} on registers A, B with orthonormal states and quantum systems A′, B′, cut AA′ : BB′ → flaggedState p ρij on (ι × a′) × (κ × b′) for finite register types ι, κ and nonempty finite a′, b′, registers the computational bases of ι, κ; ρ^ij_{A′} = traceRight (ρij ij), ρ^ij_{B′} = traceLeft (ρij ij); reason: any orthonormal register gives a unitarily equivalent state with the same entropies; empty ι or κ is excluded by Σ p = 1; same instances
Preamble
import Mathlib
import Definitions.Def_NoLocalBroadcast_Broadcast_Model
Formal statement
namespace NoLocalBroadcast.Broadcast

open Matrix
open scoped Kronecker

/-- Display (1), p. 3: for `σ̃ = Σ_{ij} p_{ij} |i⟩⟨i| ⊗ |j⟩⟨j| ⊗ ρ^{ij}` (cut `AA′ : BB′`),
`I(σ̃) = I({p_ij}) + Σ_i p_i^A S(τ^i_{A′}) + Σ_j p_j^B S(τ^j_{B′}) − Σ_{ij} p_{ij} S(ρ^{ij})`
and `I(σ̃) ≥ I({p_ij}) + Σ_{ij} p_{ij} I(ρ^{ij})`, where `I({p_ij})` is the mutual information
of the diagonal state of `p`. -/
theorem display_1 {ι κ a' b' : Type} [Fintype ι] [DecidableEq ι] [Fintype κ] [DecidableEq κ]
    [Fintype a'] [DecidableEq a'] [Nonempty a'] [Fintype b'] [DecidableEq b'] [Nonempty b']
    (p : ι × κ → ℝ) (hp0 : ∀ ij, 0 ≤ p ij) (hp1 : ∑ ij, p ij = 1)
    (ρij : ι × κ → Matrix (a' × b') (a' × b') ℂ)
    (hρij : ∀ ij, QAlgorithms.IsDensityMatrix (ρij ij)) :
    mutualInfo (flaggedState p ρij) =
        mutualInfo (Matrix.diagonal fun ij => (p ij : ℂ))
          + ∑ i, (∑ j, p (i, j)) * QAlgorithms.vnEntropy
              (∑ j, ((p (i, j) / ∑ j', p (i, j') : ℝ) : ℂ) • QAlgorithms.traceRight (ρij (i, j)))
          + ∑ j, (∑ i, p (i, j)) * QAlgorithms.vnEntropy
              (∑ i, ((p (i, j) / ∑ i', p (i', j) : ℝ) : ℂ) • QAlgorithms.traceLeft (ρij (i, j)))
          - ∑ ij, p ij * QAlgorithms.vnEntropy (ρij ij) ∧
    mutualInfo (Matrix.diagonal fun ij => (p ij : ℂ)) + ∑ ij, p ij * mutualInfo (ρij ij)
      ≤ mutualInfo (flaggedState p ρij) := by sorry

end NoLocalBroadcast.Broadcast
Source
Piani, Horodecki & Horodecki, No-local-broadcasting theorem for quantum correlations, arXiv:0707.0848v2, p. 3, display (1) in the proof of Theorem 3
Human review
  • Endorsed by Shuze Chen · Oct 24, 2026

    Confirmed by the moderator at approval.

  • Endorsed by mikedeng1 · 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