Lemma 11 lower bounds (B6), (B8)
OpenQAlgorithms.Papers.Independent.Intakenext100.P20.schur_block_lower_boundsLet and let be a normalized density matrix with decreasing spectrum , allowing zeros. For a Young index , let be the orthogonal projector onto its full Schur-Weyl isotypic block and let . Write for extended classical relative entropy, and put . For every ,
For any subset of the probability simplex, let and . Then
If is nonempty, the finite minimum is attained, including when every value is infinite. If is empty, and the left side is zero. Positive empirical mass outside the spectral support also gives a zero exponential. These are precisely the lower clauses (B6), (B8), without the defective upper-bound constants.
import Definitions.Def_QAlgorithms_Papers_Independent_Intakenext100_P20
noncomputable section
open scoped Matrix ENNReal ComplexOrder Topology
namespace QAlgorithms.Papers.Independent.Intakenext100.P20
open QAlgorithms.Papers.Independent.Information100.P08 (IsPure)
open QAlgorithms.Papers.Independent.Information100.P47 (IsProbability)
open QAlgorithms.Papers.Independent.Information100.P53 (SpectralData)
open QAlgorithms.Papers.Independent.Gap48.P19 (shannonNats)
open QAlgorithms.Papers.Independent.Dependency50.P28 (classicalD)
attribute [local instance] Classical.propDecidable
/-- Lemma 11, PDF p.10: lower bounds (B6), (B8), including attainment of
the finite type minimum when nonempty and zero contribution for an empty event. -/
theorem schur_block_lower_bounds {d n : ℕ} (hd : 0 < d) (hn : 0 < n)
(rho : {M : Matrix (Fin d) (Fin d) ℂ // IsDensityMatrix M})
(spectral : SpectralData rho.val) (A : Set (Fin d → ℝ))
(hA : ∀ b ∈ A, IsProbability b) :
(∀ shape : Schur.Partition d n,
lowerPrefactor d n * expNegD n
((classicalD (.finite 1 (by norm_num)) (Schur.normalizedRows hd hn shape) spectral.value).toENNReal) ≤ blockMass rho.val shape) ∧
lowerPrefactor d n * expNegD n (eventD hd hn spectral.value A) ≤
∑ shape ∈ eventTypes hd hn A, blockMass rho.val shape ∧
((eventTypes hd hn A).Nonempty →
∃ shape ∈ eventTypes hd hn A,
(classicalD (.finite 1 (by norm_num)) (Schur.normalizedRows hd hn shape) spectral.value).toENNReal =
eventD hd hn spectral.value A) := by sorry
end QAlgorithms.Papers.Independent.Intakenext100.P20
end
Read-back
What the Lean code literally says, in plain math · gpt-6.1-sol
For all positive integers , let be a positive semidefinite complex matrix of trace , and suppose spectral data are supplied consisting of a nonincreasing real sequence and an orthonormal basis of such that . Let be any set whose members have nonnegative coordinates summing to . The shapes considered are all Young diagrams with exactly cells and at most rows; write , where is the length of row , including zero lengths. Define as the extended nonnegative value obtained from when every positive has positive , and as otherwise; the conversion to an extended nonnegative value sends negative values to . Here is the natural logarithm. Put for finite , and , and put . Define the mass , where , and is the following explicitly specified matrix. Using the fixed numbering of the cells of by , let consist respectively of permutations preserving each cell’s row and column, let have entries , let , and let be the orthogonal projector onto this range. With , , and , set . Finally, let and , with the infimum taken in . The declaration asserts all three conclusions:
and, if , there exists with . Empty event sets are included: their infimum is , their mass sum is , and the attainment conclusion is conditional on nonemptiness.
Confirmed by the mission captain (proposal self-audit).
Confirmed by the moderator at approval.