Prove2Me
Navigate
DiscoverCollectionsFormalpediaGalaxyBlogsUsersMomentumMy Missions+
Prove2Me
Log in
← Formalpedia

Proposition V.2: average monotonicity under LOCC

Open
QAlgorithms.Papers.Independent.Support50.P09.squashed_LOCC_average

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

prove2quantum

For a normalized state ρX1⋯Xm\rho^{X_1\cdots X_m}ρX1​⋯Xm​ with m≥1m\ge1m≥1 finite nonzero local systems, define

Esq(ρ)=12inf⁡ρ~:Tr⁡Eρ~=ρ[∑i=1mH(Xi∣E)ρ~−H(X1⋯Xm∣E)ρ~].E_{\rm sq}(\rho)=\frac12\inf_{\widetilde\rho:\operatorname{Tr}_E\widetilde\rho=\rho} \left[\sum_{i=1}^m H(X_i\mid E)_{\widetilde\rho}-H(X_1\cdots X_m\mid E)_{\widetilde\rho}\right].Esq​(ρ)=21​ρ​:TrE​ρ​=ρinf​[i=1∑m​H(Xi​∣E)ρ​​−H(X1​⋯Xm​∣E)ρ​​].

Entropies are in bits. The infimum ranges over all normalized extensions and every positive finite dimension of EEE, with no dimension bound or attainment assumption. One-party inputs use this same entropy expression and have value zero.

Let a finite adaptive LOCC protocol applied to ρ\rhoρ have terminal normalized states ρj\rho_jρj​ and probabilities pjp_jpj​. Local dimensions may change with the branch, and classical transcripts may be forgotten. Then

∑jpjEsq(ρj)≤Esq(ρ).\sum_jp_jE_{\rm sq}(\rho_j)\le E_{\rm sq}(\rho).j∑​pj​Esq​(ρj​)≤Esq​(ρ).

Zero-probability branches use arbitrary normalized states with the subnormalized equation pjρj=Ej(ρ)p_j\rho_j=\mathcal E_j(\rho)pj​ρj​=Ej​(ρ), without dividing by pjp_jpj​.

This retains the proposition’s average monotonicity, including selective and deterministic finite-round operations.

Preamble
import Definitions.Def_QAlgorithms_Papers_Independent_Support50_P09
Formal statement
noncomputable section

namespace QAlgorithms.Papers.Independent.Support50.P09

open QAlgorithms.Papers.Independent.Claimed50.P33

open scoped Matrix ComplexConjugate ComplexOrder

attribute [local instance] Classical.propDecidable Classical.decEq

/-- Proposition V.2, PDF pp.16–17: AVERAGE monotonicity for finite adaptive
LOCC, including branch-dependent dimensions and forgotten transcripts.
Zero-probability outcomes use subnormalized equations, not division by zero. -/
theorem squashed_LOCC_average {m : ℕ} (hm : 0 < m) :
    IsAdaptiveLOCCMonotone (fun s : FiniteState m => halfExtensionSquashed .total s.rho) := by sorry

end QAlgorithms.Papers.Independent.Support50.P09

end
Source
Distributed Compression and Multiparty Squashed Entanglement, arXiv:0707.2792v2, PDF p. 16, Proposition V.2: average monotonicity under LOCC
Read-back

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

For every positive natural number mmm, consider states sss consisting of positive integer local dimensions and a positive semidefinite matrix ρs\rho_sρs​ of trace 111. For any such mmm-party density matrix, define

Em(ρ)=12inf⁡e>0, σ a density matrixTr⁡Eσ=ρ[∑i=0m−1S(σiE)−S(σ)−(m−1)S(σE)],E_m(\rho)=\frac12\inf_{\substack{e>0,\ \sigma\ \text{a density matrix}\\ \operatorname{Tr}_{E}\sigma=\rho}} \left[\sum_{i=0}^{m-1}S(\sigma_{iE})-S(\sigma)-(m-1)S(\sigma_E)\right],Em​(ρ)=21​e>0, σ a density matrixTrE​σ=ρ​inf​[i=0∑m−1​S(σiE​)−S(σ)−(m−1)S(σE​)],

where eee ranges over positive integer environment dimensions, and S(X)=−Re⁡tr⁡(f(X))S(X)=-\operatorname{Re}\operatorname{tr}(f(X))S(X)=−Retr(f(X)) for f(t)=tlog⁡2tf(t)=t\log_2tf(t)=tlog2​t, with f(0)=0f(0)=0f(0)=0. For every initial state sss and every finite list of weighted output states (wℓ,sℓ)(w_\ell,s_\ell)(wℓ​,sℓ​) obtainable by the following inductive rules, one has

∑ℓwℓEm(ρsℓ)≤Em(ρs).\sum_\ell w_\ell E_m(\rho_{s_\ell})\le E_m(\rho_s).ℓ∑​wℓ​Em​(ρsℓ​​)≤Em​(ρs​).

The rules allow stopping with the single output (1,s)(1,s)(1,s); applying a finite local instrument and then choosing a further such procedure separately for each outcome; and replacing an output list whose states all have the same local dimensions by its weighted mixture, with weight 111. A permitted local instrument has some party p<mp<mp<m, finitely many outcomes j<kj<kj<k, weights pj≥0p_j\ge0pj​≥0 summing to 111, and density matrices ρrj\rho_{r_j}ρrj​​; all dimensions other than party ppp's agree with those of the input. Each outcome has a finite family of matrices Aj,zA_{j,z}Aj,z​ from the input space of party ppp to its output space. Writing A~j,z\widetilde A_{j,z}Aj,z​ for their action on party ppp with identity action on every other party, they satisfy ∑j,zA~j,z†A~j,z=I\sum_{j,z}\widetilde A_{j,z}^{\dagger}\widetilde A_{j,z}=I∑j,z​Aj,z†​Aj,z​=I and pjρrj=∑zA~j,zρsA~j,z†p_j\rho_{r_j}=\sum_z\widetilde A_{j,z}\rho_s\widetilde A_{j,z}^{\dagger}pj​ρrj​​=∑z​Aj,z​ρs​Aj,z†​. Branch weights multiply along successive instruments. The mixture rule requires equality of all matrix entries with the weighted sum. Zero-weight branches and m=1m=1m=1 are included.

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