Convex Petz moment, entropy slope and compact maximizers
ProvedQAlgorithms.Papers.Independent.Gap48.P41.petz_moment_propertiesLet be a nonempty finite orthonormal basis, and let be positive semidefinite complex matrices of trace one with . For a positive integer , an arbitrary effect accepts the null state . Write and . All logarithms are natural, with in rates. Write , using support-restricted spectral logarithms inside the trace.
Put and . Spectral powers vanish on the kernel, including at zero and negative orders.
The function is convex on , , and . Moreover,
For and ,
These properties identify the entropy threshold and the compact optimizer condition, without claiming uniqueness.
Formalization Note Endpoint derivatives use the finite spectral expression on a neighborhood of the interval. The displayed strict second-derivative claim on page 3 is corrected to nonnegative curvature, as required by the equal-state case.
import Definitions.Def_QAlgorithms_Papers_Independent_Gap48_P41
noncomputable section
namespace QAlgorithms.Papers.Independent.Gap48.P41
open scoped Matrix ComplexOrder BigOperators
open Filter QAlgorithms
open QAlgorithms.Papers.Independent.Information100.P60 (Mat copies)
open QAlgorithms.Papers.Independent.Information100.P71 (supportPower)
open QAlgorithms.Papers.Independent.Information100.P07 (positiveProjection)
open QAlgorithms.Papers.Independent.Dependency50.P28
(tr SupportLE relativeEntropy scalarPower classicalD typeI accept TestSequence
scaled logNonnegativeExtended)
open QAlgorithms.Papers.Independent.Claimed50.P15 (legendre)
open QAlgorithms.Papers.Independent.Information100.P56 (IsLaw)
attribute [local instance] Classical.propDecidable Classical.decEq
variable {I : Type} [Fintype I] [DecidableEq I]
/-- PDF pp.3--4, unnumbered properties and (9): convexity, slope, positivity,
and membership in the set of maximizers. SOURCE CORRECTION: the displayed
psi''>0 on p.3 must be psi''>=0; identical states give psi=0. The prose convexity
claim is retained without an invented nondegeneracy or unique-maximizer premise. -/
theorem petz_moment_properties [Nonempty I]
(rho sigma : Mat I) (hrho : IsDensityMatrix rho) (hsigma : IsDensityMatrix sigma)
(hsupp : SupportLE rho sigma) :
ConvexOn ℝ (Set.Icc 0 1) (petzPsi rho sigma) ∧
petzPsi rho sigma 0 = 0 ∧
HasDerivAt (petzPsi rho sigma) (relativeEntropy rho sigma).toReal 0 ∧
(∀ lambda : ℝ, (relativeEntropy rho sigma).toReal < lambda →
0 < petzPhi rho sigma lambda) ∧
(∀ lambda : ℝ, (relativeEntropy rho sigma).toReal ≤ lambda →
lambda ≤ deriv (petzPsi rho sigma) 1 → ∀ s ∈ Set.Icc (0 : ℝ) 1,
(IsGreatest ((fun t : ℝ => lambda * t - petzPsi rho sigma t) '' Set.Icc 0 1)
(lambda * s - petzPsi rho sigma s) ↔ deriv (petzPsi rho sigma) s = lambda)) := by sorry
end QAlgorithms.Papers.Independent.Gap48.P41
end
Read-back
What the Lean code literally says, in plain math · gpt-6.1-sol
Let be a nonempty finite type with decidable equality, and let be positive semidefinite complex matrices indexed by , each of trace , with . Define by spectral functional calculus using the function that is at zero and otherwise, and set
The matrix logarithms use spectral functional calculus with the total real logarithm, for which ; the displayed denominator equals . Then is convex on , , and its ordinary derivative at exists and equals . Every real satisfies . Moreover, for every real satisfying and every , including either endpoint, the value is the greatest member of if and only if . No assumption that is imposed.
Confirmed by the mission captain (proposal self-audit).
Confirmed by the moderator at approval.