Gaussian coherent-state mixtures and observed quadrature broadening
OpenQAlgorithms.Papers.Independent.Intakethird100.P02.gaussianBroadeningLet , and displace a coherent amplitude by an independent centered isotropic Gaussian variable with planar density . Use coherent quadrature variance . The coherent-state mixture
has integrable number-basis coefficients and is a positive trace-class operator of trace one. Either homodyne quadrature has a Gaussian law, mean equal to the corresponding quadrature of , finite second moment, and variance
For equally likely real signal amplitudes with , the observed density is the equal mixture of real Gaussians with means and variance . Its posterior and conditional Shannon information are
This identifies the variance substitution in Eqs.(2) and (5). It makes no claim that Eve's information remains independent of the homodyne outcome under general noise.
import Definitions.Def_QAlgorithms_Papers_Independent_Intakethird100_P02
noncomputable section
namespace QAlgorithms.Papers.Independent.Intakethird100.P02
open MeasureTheory ProbabilityTheory
open scoped ENNReal Topology
open QAlgorithms.Papers.Independent.Dependency50.P39
(Fock FockOperator coherentKet coherentCoefficient)
open QAlgorithms.Papers.Independent.Frontier30.P21 (coherentDensity)
open QAlgorithms.Papers.Independent.Next46.P32 (complexNormal quadrature)
open QAlgorithms.Papers.Independent.Dependency50.P33 (IsDensity productCoeff)
/-- PDF p.3, (10) and the following paragraph: Gaussian convolution gives
variance 1/4+1/(2 lambda). The observed binary density and Bayes posterior
use this variance. No assertion about Eve's information under general noise. -/
theorem gaussianBroadening (lambda : ℝ) (hlambda : 0 < lambda) :
(∀ (alpha : ℂ) (k : Bool),
MixtureIntegrable alpha (1 / lambda) ∧
IsDensity (displacedCoherent alpha (1 / lambda)) ∧
displacedHomodyne alpha (1 / lambda) k =
gaussianReal (coherentAmplitudeQuadrature k alpha) ⟨(1 / 4 : ℝ) + 1 / (2 * lambda), by
positivity⟩ ∧
MemLp (fun x : ℝ => x) 2 (displacedHomodyne alpha (1 / lambda) k) ∧
variance (fun x : ℝ => x) (displacedHomodyne alpha (1 / lambda) k) =
(1 / 4 : ℝ) + 1 / (2 * lambda)) ∧
(∀ energy : ℝ, 0 < energy →
binaryDisplacedLaw (1 / lambda) energy = volume.withDensity
(fun x => ENNReal.ofReal
(binaryDensity ((1 / 4 : ℝ) + 1 / (2 * lambda)) energy x))) ∧
(∀ (energy x : ℝ), 0 < energy →
homodynePosterior ((1 / 4 : ℝ) + 1 / (2 * lambda)) energy x =
1 / (1 + Real.exp (-2 * Real.sqrt energy * x /
((1 / 4 : ℝ) + 1 / (2 * lambda)))) ∧
homodyneInformation ((1 / 4 : ℝ) + 1 / (2 * lambda)) energy x =
binaryInformation (1 / (1 + Real.exp (-2 * Real.sqrt energy * x /
((1 / 4 : ℝ) + 1 / (2 * lambda)))))) := by sorry
end QAlgorithms.Papers.Independent.Intakethird100.P02
end
Read-back
What the Lean code literally says, in plain math · gpt-6.1-sol
For every real , put and , and let be the complex Gaussian measure with planar density . For every complex and Boolean , all functions , for , are -integrable, where ; the matrix is a density in the following precise sense: there exist sequences with and ; every finite quadratic form , for a finite and arbitrary complex sequence , is real and nonnegative; and is summable and equals one. Moreover, the pushforward of times a real mean-zero Gaussian of variance under , where and , is the real Gaussian measure of mean and variance ; its identity function belongs to and has variance exactly . For every real , the equal mixture of these pushforwards with real means has Lebesgue density
where is the Gaussian density of variance . Finally, for every real and every real ,
and the quantity evaluated at the density-ratio posterior equals that same expression evaluated at . The assertions include both quadrature choices, all natural matrix indices including zero, and every real ; the assertions concerning require strict positivity.
Confirmed by the mission captain (proposal self-audit).
Confirmed by the moderator at approval.