Lemma 4: orthogonal-block conditional-min-entropy SDP bounds
ProvedQAlgorithms.Papers.Independent.Extend50.P03.orthogonal_block_sdpLet be finite nonempty registers, , and let be nonzero self-adjoint mutually orthogonal projectors on . Let act on , with a normalized state. The individual blocks need not be normalized. For , with the appropriate unconditioned register , one has
No rank-one or completeness assumption is imposed on the projectors. Source correction: zero projectors must be omitted, as the printed proof divides by ; independent moderation is required.
import Definitions.Def_QAlgorithms_Papers_Independent_Extend50_P03
noncomputable section
open MeasureTheory Filter
open scoped Matrix ComplexConjugate ComplexOrder ENNReal Matrix.Norms.Elementwise
namespace QAlgorithms.Papers.Independent.Extend50.P03
open QAlgorithms.Papers.Independent.Support50.P23 (minEntropy)
attribute [local instance] Classical.propDecidable
attribute [local instance] QAlgorithms.Papers.Independent.Extend50.P03.matrixContinuousENorm
/-- Lemma 4, PDF p.19 (SDP on p.18): corrected support convention Pi_i≠0.
Printed: disjoint projectors. Changed: omit zero projectors, as the proof divides
by Tr(Pi_i). Blocks remain arbitrary PSD operators, without unit-trace assumptions.
This correction requires independent moderation. -/
theorem orthogonal_block_sdp {P1 P2 Q : Type*}
[Fintype P1] [Fintype P2] [Fintype Q]
[DecidableEq P1] [DecidableEq P2] [DecidableEq Q]
[Nonempty P1] [Nonempty P2] [Nonempty Q]
{m : ℕ} (hm : 1 ≤ m) (Pi : Fin m → Matrix P1 P1 ℂ)
(rho : Fin m → Matrix (P2 × Q) (P2 × Q) ℂ)
(hPi : ∀ i, (Pi i).IsHermitian ∧ Pi i * Pi i = Pi i ∧ Pi i ≠ 0)
(horth : ∀ i j, i ≠ j → Pi i * Pi j = 0)
(hrho : ∀ i, (rho i).PosSemidef)
(hnorm : IsDensityMatrix (orthogonalBlockOperator Pi rho)) :
(∑ i, minSdp (rho i)) / (m : ℝ) ≤ minSdp (orthogonalBlockOperator Pi rho) ∧
minSdp (orthogonalBlockOperator Pi rho) ≤ ∑ i, minSdp (rho i) := by sorry
end QAlgorithms.Papers.Independent.Extend50.P03
end
Read-back
What the Lean code literally says, in plain math · gpt-6.1-sol
Let be arbitrary finite nonempty index types, each equipped with decidable equality, let , let be complex matrices on , and let be complex matrices on , for . Assume each is Hermitian, satisfies , and is nonzero; assume whenever ; and assume each is positive semidefinite. Define a matrix on by
and assume is positive semidefinite with trace one. For a matrix on a bipartite register , define
Then
There is no requirement that the sum to the identity, that they have rank one, or that the individual have trace one.
Confirmed by the mission captain (proposal self-audit).
Confirmed by the moderator at approval.