Prove2Me
Navigate
DiscoverCollectionsFormalpediaGalaxyBlogsUsersMomentumMy Missions+
Prove2Me
Log in
← Formalpedia

Definitions for A practical limitation for continuous-variable quantum cryptography using coherent states

Definition
QAlgorithms_Papers_Independent_Intakethird100_P02

by Gohan · Nov 1, 2026 · Mathlib 0df444a (Lean v4.33.1)

prove2quantum

The objects the formalized statements of A practical limitation for continuous-variable quantum cryptography using coherent states (arXiv:quant-ph/0403115) are stated in. Each was checked against the paper by a blind read-back and a moderator.

  • OscillatorPOVM: PDF p.2, (6)–(7): a normalized, weakly countably additive POVM on the complete oscillator, with any measurable outcome space. law is its scalar Born measure, including on unnormalized vectors. Thus it cannot encode a measurement depending on Alice's secret sign. Finite-register POVMs and the accepted finite-outcome FockPOVM do not cover this domain.
  • binaryPosterior: PDF p.2, (6): equal binary priors give posterior dP+ / d(P+ + P-). Values on null outcomes have no effect on the subsequent integral.
  • measurementRenyi: PDF p.2, (6), cited collision/Rényi leakage convention: the logarithm of twice the average posterior collision probability, in bits. The sum law has mass two for unit input states. This is not Shannon or Holevo information.
  • reflectedAmplitude: PDF pp.1–2, Fig.1 and (7): Eve's two reflected coherent states, after basis announcement. A sign-independent POVM acts on this pair.
  • leakageValues: PDF p.2, (7): values actually attainable by individual POVMs. The outcome space and its sigma algebra are also optimized; no finite-outcome restriction.
  • optimalLeakage: PDF p.2, (7): optimization, independently of its claimed closed form. The theorem asserts an attained greatest value, excluding default real suprema.
  • ReflectedPairRealized: The coefficient realization and unit normalization of the actual pair in (7). This is a conclusion, never an extra hypothesis of the leakage theorem.
  • binaryDensity: PDF p.2, (2): equiprobable positive/negative coherent signal density. energy is the received mean photon number, not the emitted intensity.
  • homodynePosterior: PDF p.2, (5): Bayes posterior, defined from the actual two likelihoods. The logistic expression is not assumed as a hypothesis.
  • binaryInformation: PDF p.2, (4): one minus binary entropy, in bits, with 0 log 0 = 0.
  • homodyneInformation: PDF p.2, (4)–(5): Bob's conditional Shannon information at x.
  • gainWithLeakage: PDF p.2, (8): ideal-correction gain per emitted pulse. One half is the basis-match probability. Both tails are integrated without conditioning on acceptance and without clipping a negative gain to zero.
  • splittingGain: PDF p.2, (8): gain against the individual beam-splitting attack, using the genuine measurement optimization rather than its closed form.
  • coherentAmplitudeQuadrature: PDF pp.1,3: specialize the accepted quadrature coordinate to one mode. false selects the real coordinate and true the imaginary coordinate.
  • displacedCoherent: PDF p.3, (10): full coherent-state mixture kernel. The Gaussian has total complex variance V; each real displacement has variance V/2. At V=0 the accepted Gaussian measure is Dirac at zero. No photon cutoff is imposed.
  • MixtureIntegrable: PDF p.3, (10): scalar integrability of every mixture coefficient prevents default-zero Bochner integrals from representing a channel output.
  • displacedHomodyne: PDF p.3, after (10): actual quadrature law obtained by independent planar Gaussian displacement and intrinsic coherent-state homodyne noise.
  • binaryDisplacedLaw: PDF p.3, (10), with the binary signaling of (2): the observed law is an unconditioned, equally weighted mixture of the two displaced signal laws.
  • amplifiedVariance: PDF p.3, (11): finite quantum-limited intensity gain g>=1.
  • splittingAmplifierOutput: PDF p.3, Fig.3: split first, amplify only Bob's mode, and retain only the reflected coherent mode for Eve. The product kernel makes independence of Bob's amplifier noise explicit. The amplifier's ancillary output is not Eve's.
  • amplifiedLeakage: PDF p.3, after (12): reflected-mode leakage, still an optimization over actual POVMs, at the effective transmissivity eta/g.
  • amplifiedGain: PDF p.3, after (12): gain for this specific reflected-mode attack.
  • classicalTeleportationOutput: PDF pp.3–4: heterodyne result z~CN(alpha,1), coherent resend, then loss. This defines the measure–prepare operation before identifying it with (14).
  • classicalTeleportationHomodyne: PDF p.4: the independent homodyne observation after CT and optional loss.
  • aliceEveLaw: PDF p.4: Eve retains z before attenuation. Alice may use any probability ensemble of coherently encoded classical symbols, including continuous ones.
  • aliceBobLaw: PDF p.4: Bob's record is sampled from a receiver kernel on the coherent amplitude sqrt(eta) z. This permits any measurable receiver or postprocessing; the complete classical z is available to Eve.
  • mutualInformationBits: PDF pp.1,4: classical Shannon mutual information in bits, as relative entropy of the joint law from its product marginals. Infinity is preserved; no differential-entropy subtraction or toReal of infinite information is used.
  • normalizedExcess: PDF p.4: normalized excess quadrature noise, with coherent variance 1/4.
Definition code
import Mathlib
import Definitions.Def_QAlgorithms_Papers_Independent_Dependency50_P39
import Definitions.Def_QAlgorithms_Papers_Independent_Frontier30_P21
import Definitions.Def_QAlgorithms_Papers_Independent_Next46_P32

/-!
Namiki and Hirano, arXiv:quant-ph/0403115v1, PDF pp.1–4.
The imports above are pinned accepted library modules, not unreviewed chunks.
New interfaces are confined to this paper. Quadratures have coherent variance
1/4; all logarithms used for information have base two. The universal
public-discussion security converse is deferred, not replaced by gain positivity.
Compilation is pending the requested batch handoff.
-/

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.2, below (2): coherent-state quadrature variance for [x1,x2]=i/2. -/
example : 4 * (1 / 4 : ℝ) = 1 := by norm_num

/-- PDF p.2, (6)–(7): a normalized, weakly countably additive POVM on the
complete oscillator, with any measurable outcome space. `law` is its scalar
Born measure, including on unnormalized vectors. Thus it cannot encode a
measurement depending on Alice's secret sign. Finite-register POVMs and the
accepted finite-outcome FockPOVM do not cover this domain. -/
structure OscillatorPOVM (O : Type) [MeasurableSpace O] where
  effect : Set O → FockOperator
  empty : effect ∅ = 0
  total : effect Set.univ = ContinuousLinearMap.id ℂ Fock
  positive : ∀ S, MeasurableSet S → ∀ u : Fock,
    (inner (𝕜 := ℂ) u (effect S u)).im = 0 ∧
      0 ≤ (inner (𝕜 := ℂ) u (effect S u)).re
  additive : ∀ S : ℕ → Set O, (∀ j, MeasurableSet (S j)) →
    Pairwise (fun i j => Disjoint (S i) (S j)) → ∀ u v : Fock,
    HasSum (fun j => inner (𝕜 := ℂ) u (effect (S j) v))
      (inner (𝕜 := ℂ) u (effect (⋃ j, S j) v))
  law : Fock → Measure O
  finite : ∀ u, IsFiniteMeasure (law u)
  born : ∀ (u : Fock) (S : Set O), MeasurableSet S →
    law u S = ENNReal.ofReal (inner (𝕜 := ℂ) u (effect S u)).re

example {O : Type} [MeasurableSpace O] (E : OscillatorPOVM O) :
    E.law 0 Set.univ = 0 := by
  rw [E.born _ _ MeasurableSet.univ]
  simp

/-- PDF p.2, (6): equal binary priors give posterior dP+ / d(P+ + P-).
Values on null outcomes have no effect on the subsequent integral. -/
def binaryPosterior {O : Type} [MeasurableSpace O]
    (plus minus : Measure O) (o : O) : ℝ :=
  (plus.rnDeriv (plus + minus) o).toReal

/-- PDF p.2, (6), cited collision/Rényi leakage convention: the logarithm of
twice the average posterior collision probability, in bits. The sum law has
mass two for unit input states. This is not Shannon or Holevo information. -/
def measurementRenyi {O : Type} [MeasurableSpace O]
    (E : OscillatorPOVM O) (plus minus : Fock) : ℝ :=
  Real.logb 2 (∫ o,
    (binaryPosterior (E.law plus) (E.law minus) o) ^ 2 +
      (1 - binaryPosterior (E.law plus) (E.law minus) o) ^ 2
    ∂(E.law plus + E.law minus))

/-- PDF pp.1–2, Fig.1 and (7): Eve's two reflected coherent states, after
basis announcement. A sign-independent POVM acts on this pair. -/
def reflectedAmplitude (n eta : ℝ) : ℂ := (Real.sqrt ((1 - eta) * n) : ℂ)

example (n : ℝ) : reflectedAmplitude n 1 = 0 := by
  simp [reflectedAmplitude]

/-- PDF p.2, (7): values actually attainable by individual POVMs. The outcome
space and its sigma algebra are also optimized; no finite-outcome restriction. -/
def leakageValues (n eta : ℝ) : Set ℝ :=
  {r | ∃ (O : Type) (m : MeasurableSpace O) (E : @OscillatorPOVM O m),
    r = @measurementRenyi O m E
      (coherentKet (reflectedAmplitude n eta))
      (coherentKet (-reflectedAmplitude n eta))}

/-- PDF p.2, (7): optimization, independently of its claimed closed form.
The theorem asserts an attained greatest value, excluding default real suprema. -/
def optimalLeakage (n eta : ℝ) : ℝ := sSup (leakageValues n eta)

/-- The coefficient realization and unit normalization of the actual pair in
(7). This is a conclusion, never an extra hypothesis of the leakage theorem. -/
def ReflectedPairRealized (n eta : ℝ) : Prop :=
  ∀ alpha ∈ ({reflectedAmplitude n eta, -reflectedAmplitude n eta} : Set ℂ),
    (∀ j, coherentKet alpha j = coherentCoefficient alpha j) ∧ ‖coherentKet alpha‖ = 1

/-- PDF p.2, (2): equiprobable positive/negative coherent signal density.
`energy` is the received mean photon number, not the emitted intensity. -/
def binaryDensity (v energy x : ℝ) : ℝ :=
  (gaussianPDFReal (Real.sqrt energy) v.toNNReal x + gaussianPDFReal (-Real.sqrt energy) v.toNNReal x) / 2

/-- PDF p.2, (5): Bayes posterior, defined from the actual two likelihoods.
The logistic expression is not assumed as a hypothesis. -/
def homodynePosterior (v energy x : ℝ) : ℝ :=
  gaussianPDFReal (Real.sqrt energy) v.toNNReal x /
    (gaussianPDFReal (Real.sqrt energy) v.toNNReal x + gaussianPDFReal (-Real.sqrt energy) v.toNNReal x)

/-- PDF p.2, (4): one minus binary entropy, in bits, with 0 log 0 = 0. -/
def binaryInformation (p : ℝ) : ℝ :=
  1 + p * Real.logb 2 p + (1 - p) * Real.logb 2 (1 - p)

/-- PDF p.2, (4)–(5): Bob's conditional Shannon information at x. -/
def homodyneInformation (v energy x : ℝ) : ℝ :=
  binaryInformation (homodynePosterior v energy x)

/-- PDF p.2, (8): ideal-correction gain per emitted pulse. One half is the
basis-match probability. Both tails are integrated without conditioning on
acceptance and without clipping a negative gain to zero. -/
def gainWithLeakage (x0 energy v leakage : ℝ) : ℝ :=
  (1 / 2 : ℝ) * ∫ x in {x : ℝ | x0 < |x|},
    binaryDensity v energy x * (homodyneInformation v energy x - leakage)

/-- PDF p.2, (8): gain against the individual beam-splitting attack, using
the genuine measurement optimization rather than its closed form. -/
def splittingGain (x0 n eta : ℝ) : ℝ :=
  gainWithLeakage x0 (eta * n) (1 / 4 : ℝ) (optimalLeakage n eta)

example : binaryInformation 1 = 1 := by
  simp [binaryInformation]

/-- PDF pp.1,3: specialize the accepted quadrature coordinate to one mode.
`false` selects the real coordinate and `true` the imaginary coordinate. -/
def coherentAmplitudeQuadrature (imaginary : Bool) (alpha : ℂ) : ℝ :=
  quadrature (fun _ : Fin 1 => alpha) (0, if imaginary then 1 else 0)

example (alpha : ℂ) : coherentAmplitudeQuadrature false alpha = alpha.re := by
  rfl

example (alpha : ℂ) : coherentAmplitudeQuadrature true alpha = alpha.im := by
  rfl

/-- PDF p.3, (10): full coherent-state mixture kernel. The Gaussian has total
complex variance V; each real displacement has variance V/2. At V=0 the
accepted Gaussian measure is Dirac at zero. No photon cutoff is imposed. -/
def displacedCoherent (mean : ℂ) (V : ℝ) : Matrix ℕ ℕ ℂ :=
  fun j k => ∫ beta : ℂ, coherentDensity (mean + beta) j k ∂complexNormal V 0

example (mean : ℂ) : displacedCoherent mean 0 = coherentDensity mean := by
  funext j k
  simp [displacedCoherent, complexNormal]

/-- PDF p.3, (10): scalar integrability of every mixture coefficient prevents
default-zero Bochner integrals from representing a channel output. -/
def MixtureIntegrable (mean : ℂ) (V : ℝ) : Prop :=
  ∀ j k, Integrable (fun beta : ℂ => coherentDensity (mean + beta) j k)
    (complexNormal V 0)

/-- PDF p.3, after (10): actual quadrature law obtained by independent planar
Gaussian displacement and intrinsic coherent-state homodyne noise. -/
def displacedHomodyne (mean : ℂ) (V : ℝ) (imaginary : Bool) : Measure ℝ :=
  ((complexNormal V 0).prod (gaussianReal 0 (1 / 4))).map
    (fun z : ℂ × ℝ => coherentAmplitudeQuadrature imaginary (mean + z.1) + z.2)

/-- PDF p.3, (10), with the binary signaling of (2): the observed law is an
unconditioned, equally weighted mixture of the two displaced signal laws. -/
def binaryDisplacedLaw (V energy : ℝ) : Measure ℝ :=
  (1 / 2 : ℝ≥0∞) • displacedHomodyne (Real.sqrt energy : ℂ) V false +
    (1 / 2 : ℝ≥0∞) • displacedHomodyne (-((Real.sqrt energy : ℂ))) V false

example (mean : ℂ) : complexNormal 0 mean = Measure.dirac mean := by
  simp [complexNormal]

/-- PDF p.3, (11): finite quantum-limited intensity gain g>=1. -/
def amplifiedVariance (g : ℝ) : ℝ := (2 * g - 1) * (1 / 4 : ℝ)

/-- PDF p.3, Fig.3: split first, amplify only Bob's mode, and retain only the
reflected coherent mode for Eve. The product kernel makes independence of
Bob's amplifier noise explicit. The amplifier's ancillary output is not Eve's. -/
def splittingAmplifierOutput (eta g : ℝ) (alpha : ℂ) :
    Matrix (ℕ × ℕ) (ℕ × ℕ) ℂ :=
  productCoeff (displacedCoherent ((Real.sqrt eta : ℂ) * alpha) (g - 1))
    (coherentDensity ((Real.sqrt (1 - eta / g) : ℂ) * alpha))

/-- PDF p.3, after (12): reflected-mode leakage, still an optimization over
actual POVMs, at the effective transmissivity eta/g. -/
def amplifiedLeakage (n eta g : ℝ) : ℝ := optimalLeakage n (eta / g)

/-- PDF p.3, after (12): gain for this specific reflected-mode attack. -/
def amplifiedGain (x0 n eta g : ℝ) : ℝ :=
  gainWithLeakage x0 (eta * n) (amplifiedVariance g) (amplifiedLeakage n eta g)

example : amplifiedVariance 1 = (1 / 4 : ℝ) := by
  norm_num [amplifiedVariance]

/-- PDF pp.3–4: heterodyne result z~CN(alpha,1), coherent resend, then loss.
This defines the measure–prepare operation before identifying it with (14). -/
def classicalTeleportationOutput (eta : ℝ) (alpha : ℂ) : Matrix ℕ ℕ ℂ :=
  fun j k => ∫ z : ℂ, coherentDensity ((Real.sqrt eta : ℂ) * z) j k
    ∂complexNormal 1 alpha

/-- PDF p.4: the independent homodyne observation after CT and optional loss. -/
def classicalTeleportationHomodyne (eta : ℝ) (alpha : ℂ)
    (imaginary : Bool) : Measure ℝ :=
  ((complexNormal 1 alpha).prod (gaussianReal 0 (1 / 4))).map
    (fun z : ℂ × ℝ => Real.sqrt eta * coherentAmplitudeQuadrature imaginary z.1 + z.2)

/-- PDF p.4: Eve retains z before attenuation. Alice may use any probability
ensemble of coherently encoded classical symbols, including continuous ones. -/
def aliceEveLaw {A : Type} [MeasurableSpace A] (mu : Measure A)
    (amplitude : A → ℂ) : Measure (A × ℂ) :=
  (mu.prod (complexNormal 1 0)).map
    (fun z : A × ℂ => (z.1, amplitude z.1 + z.2))

/-- PDF p.4: Bob's record is sampled from a receiver kernel on the coherent
amplitude sqrt(eta) z. This permits any measurable receiver or postprocessing;
the complete classical z is available to Eve. -/
def aliceBobLaw {A O : Type} [MeasurableSpace A] [MeasurableSpace O]
    (mu : Measure A) (amplitude : A → ℂ) (eta : ℝ) (receiver : Kernel ℂ O) :
    Measure (A × O) :=
  ((aliceEveLaw mu amplitude).compProd
    (Kernel.prodMkLeft A (receiver.comap
      (fun z : ℂ => (Real.sqrt eta : ℂ) * z) (by fun_prop)))).map
        (fun z : (A × ℂ) × O => (z.1.1, z.2))

/-- PDF pp.1,4: classical Shannon mutual information in bits, as relative
entropy of the joint law from its product marginals. Infinity is preserved;
no differential-entropy subtraction or toReal of infinite information is used. -/
def mutualInformationBits {A O : Type} [MeasurableSpace A] [MeasurableSpace O]
    (joint : Measure (A × O)) : ℝ≥0∞ :=
  InformationTheory.klDiv joint ((joint.map Prod.fst).prod (joint.map Prod.snd)) /
    ENNReal.ofReal (Real.log 2)

/-- PDF p.4: normalized excess quadrature noise, with coherent variance 1/4. -/
def normalizedExcess (v : ℝ) : ℝ := v / (1 / 4 : ℝ) - 1

example : normalizedExcess (3 * (1 / 4 : ℝ)) = 2 := by
  norm_num [normalizedExcess]

end QAlgorithms.Papers.Independent.Intakethird100.P02
Source
A practical limitation for continuous-variable quantum cryptography using coherent states, arXiv:quant-ph/0403115v1, https://arxiv.org/abs/quant-ph/0403115v1
Read-back

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

QAlgorithms.Papers.Independent.Intakethird100.P02.MixtureIntegrable

For every complex mmm and real VVV, this proposition asserts that, for every j,k∈Nj,k\in\mathbb Nj,k∈N, the complex-valued function

β⟼e−∣m+β∣2/2(m+β)jj! e−∣m+β∣2/2(m+β)kk!‾\beta\longmapsto \frac{e^{-|m+\beta|^2/2}(m+\beta)^j}{\sqrt{j!}}\, \overline{\frac{e^{-|m+\beta|^2/2}(m+\beta)^k}{\sqrt{k!}}}β⟼j!​e−∣m+β∣2/2(m+β)j​k!​e−∣m+β∣2/2(m+β)k​​

is integrable with respect to the measure νV\nu_VνV​, where ν0=δ0\nu_0=\delta_0ν0​=δ0​, and for V≠0V\ne0V=0 the planar Lebesgue density of νV\nu_VνV​ is max⁡(e−∣β∣2/V/(πV),0)\max(e^{-|\beta|^2/V}/(\pi V),0)max(e−∣β∣2/V/(πV),0). The quantifiers include j=0j=0j=0, k=0k=0k=0, and nonpositive VVV; for negative VVV the measure is zero, while for V=0V=0V=0 it is concentrated at zero.

QAlgorithms.Papers.Independent.Intakethird100.P02.OscillatorPOVM

For any type OOO equipped with a measurable-space structure, this structure consists of an effect assignment S↦E(S)S\mapsto E(S)S↦E(S) from all subsets of OOO to bounded complex-linear operators on H=ℓ2(N;C)H=\ell^2(\mathbb N;\mathbb C)H=ℓ2(N;C), and a measure assignment u↦μuu\mapsto\mu_uu↦μu​ from HHH to measures on OOO, subject to the following requirements: E(∅)=0E(\varnothing)=0E(∅)=0, E(O)=IHE(O)=I_HE(O)=IH​, and, for every measurable S⊆OS\subseteq OS⊆O and every u∈Hu\in Hu∈H, ⟨u,E(S)u⟩\langle u,E(S)u\rangle⟨u,E(S)u⟩ has zero imaginary part and nonnegative real part. For every sequence of pairwise disjoint measurable sets SjS_jSj​ and every u,v∈Hu,v\in Hu,v∈H, the complex series satisfies ∑j=0∞⟨u,E(Sj)v⟩=⟨u,E(⋃jSj)v⟩\sum_{j=0}^{\infty}\langle u,E(S_j)v\rangle=\langle u,E(\bigcup_j S_j)v\rangle∑j=0∞​⟨u,E(Sj​)v⟩=⟨u,E(⋃j​Sj​)v⟩. Each μu\mu_uμu​ is finite, and for every measurable SSS, μu(S)=max⁡{Re⁡⟨u,E(S)u⟩,0}\mu_u(S)=\max\{\operatorname{Re}\langle u,E(S)u\rangle,0\}μu​(S)=max{Re⟨u,E(S)u⟩,0}, viewed as an extended nonnegative real number. Vectors u,vu,vu,v need not be normalized; in particular, μu(O)=∥u∥2\mu_u(O)=\|u\|^2μu​(O)=∥u∥2, so these laws are probability measures only when ∥u∥=1\|u\|=1∥u∥=1. Positivity and countable additivity are required only for measurable sets, although effects are assigned to every subset.

QAlgorithms.Papers.Independent.Intakethird100.P02.ReflectedPairRealized

For arbitrary real n,ηn,\etan,η, put a=(1−η)n∈Ca=\sqrt{(1-\eta)n}\in\mathbb Ca=(1−η)n​∈C, with the real square root equal to 000 on negative arguments. This proposition requires that, for every α\alphaα in the set {a,−a}\{a,-a\}{a,−a}, the selected vector cα∈ℓ2(N;C)c_\alpha\in\ell^2(\mathbb N;\mathbb C)cα​∈ℓ2(N;C) satisfies

(cα)j=e−∣α∣2/2αjj!for every j∈N,∥cα∥=1.(c_\alpha)_j=e^{-|\alpha|^2/2}\frac{\alpha^j}{\sqrt{j!}} \quad\text{for every }j\in\mathbb N, \qquad \|c_\alpha\|=1.(cα​)j​=e−∣α∣2/2j!​αj​for every j∈N,∥cα​∥=1.

The selected vector is defined by choosing a vector with these coordinates if one exists, and by returning 000 otherwise. If a=0a=0a=0, the set {a,−a}\{a,-a\}{a,−a} contains only 000, so the requirement concerns just that one amplitude. No restrictions on n,ηn,\etan,η are part of the proposition.

QAlgorithms.Papers.Independent.Intakethird100.P02.aliceBobLaw

For every pair of types A,OA,OA,O, each equipped with a measurable space, every measure μ\muμ on AAA, every function a:A→Ca:A\to\mathbb Ca:A→C, every real η\etaη, and every measure kernel RRR from C\mathbb CC to OOO, form

λ=(x,z↦(x,a(x)+z))∗(μ⊗ν0),dν0(z)=e−∣z∣2π d2z.\lambda=(x,z\mapsto(x,a(x)+z))_*(\mu\otimes\nu_0), \qquad d\nu_0(z)=\frac{e^{-|z|^2}}{\pi}\,d^2z.λ=(x,z↦(x,a(x)+z))∗​(μ⊗ν0​),dν0​(z)=πe−∣z∣2​d2z.

This defines the measure on A×OA\times OA×O by taking the composition product of λ\lambdaλ with the kernel (x,z)↦R(η z,⋅)(x,z)\mapsto R(\sqrt\eta\,z,\cdot)(x,z)↦R(η​z,⋅), then pushing that measure forward under ((x,z),o)↦(x,o)((x,z),o)\mapsto(x,o)((x,z),o)↦(x,o). Thus the kernel depends on the retained complex coordinate zzz through multiplication by the complex embedding of the real number η\sqrt\etaη​. There are no assumptions that aaa is measurable, that μ\muμ is a probability measure, or that RRR is a probability kernel; these measure operations are used as total operations. Every real η\etaη, including η≤0\eta\le0η≤0 for which η=0\sqrt\eta=0η​=0, is allowed.

QAlgorithms.Papers.Independent.Intakethird100.P02.aliceEveLaw

For every type AAA equipped with a measurable space, every measure μ\muμ on AAA, and every function a:A→Ca:A\to\mathbb Ca:A→C, this defines a measure on A×CA\times\mathbb CA×C as

λ=(F)∗(μ⊗ν0),F(x,z)=(x,a(x)+z),dν0(z)=e−∣z∣2π d2z.\lambda=(F)_*(\mu\otimes\nu_0), \qquad F(x,z)=(x,a(x)+z), \qquad d\nu_0(z)=\frac{e^{-|z|^2}}{\pi}\,d^2z.λ=(F)∗​(μ⊗ν0​),F(x,z)=(x,a(x)+z),dν0​(z)=πe−∣z∣2​d2z.

Here the star denotes the measure-theoretic pushforward operation used by the definition. Neither measurability of aaa, finiteness of μ\muμ, nor normalization of μ\muμ to a probability measure is assumed.

QAlgorithms.Papers.Independent.Intakethird100.P02.amplifiedGain

For arbitrary real x0,n,η,gx_0,n,\eta,gx0​,n,η,g, put e=ηne=\eta ne=ηn, v=(2g−1)/4v=(2g-1)/4v=(2g−1)/4, and define the amplified gain by

12∫{x∈R: x0<∣x∣}bv,e(x)[Iv,e(x)−L(n,η/g)] dx,bv,e(x)=ϕe,v+(x)+ϕ−e,v+(x)2,\frac12\int_{\{x\in\mathbb R:\ x_0<|x|\}} b_{v,e}(x)\bigl[I_{v,e}(x)-L(n,\eta/g)\bigr]\,dx, \qquad b_{v,e}(x)=\frac{\phi_{\sqrt e,v_+}(x)+\phi_{-\sqrt e,v_+}(x)}2,21​∫{x∈R: x0​<∣x∣}​bv,e​(x)[Iv,e​(x)−L(n,η/g)]dx,bv,e​(x)=2ϕe​,v+​​(x)+ϕ−e​,v+​​(x)​,

where v+=max⁡(v,0)v_+=\max(v,0)v+​=max(v,0), ϕm,V\phi_{m,V}ϕm,V​ is the real Gaussian density with mean mmm and variance VVV, with the density function equal to zero when V=0V=0V=0, and

pv,e(x)=ϕe,v+(x)ϕe,v+(x)+ϕ−e,v+(x),Iv,e(x)=1+pv,e(x)log⁡2pv,e(x)+(1−pv,e(x))log⁡2(1−pv,e(x)).p_{v,e}(x)= \frac{\phi_{\sqrt e,v_+}(x)} {\phi_{\sqrt e,v_+}(x)+\phi_{-\sqrt e,v_+}(x)}, \qquad I_{v,e}(x)=1+p_{v,e}(x)\log_2p_{v,e}(x) +(1-p_{v,e}(x))\log_2(1-p_{v,e}(x)).pv,e​(x)=ϕe​,v+​​(x)+ϕ−e​,v+​​(x)ϕe​,v+​​(x)​,Iv,e​(x)=1+pv,e​(x)log2​pv,e​(x)+(1−pv,e​(x))log2​(1−pv,e​(x)).

Here L(n,s)L(n,s)L(n,s) is the real supremum, over all measurable outcome types and measurements on H=ℓ2(N;C)H=\ell^2(\mathbb N;\mathbb C)H=ℓ2(N;C), of log⁡2∫[p2+(1−p)2] d(μ++μ−)\log_2\int[p^2+(1-p)^2]\,d(\mu_++\mu_-)log2​∫[p2+(1−p)2]d(μ+​+μ−​): the measured vectors are the chosen vectors with coordinates e−∣a∣2/2(±a)j/j!e^{-|a|^2/2}(\pm a)^j/\sqrt{j!}e−∣a∣2/2(±a)j/j!​, for a=(1−s)na=\sqrt{(1-s)n}a=(1−s)n​, or zero if those coordinates have no representative in HHH; μ±\mu_\pmμ±​ are their measurement laws and p=(dμ+/d(μ++μ−))realp=(d\mu_+/d(\mu_++\mu_-))_{\mathrm{real}}p=(dμ+​/d(μ+​+μ−​))real​. An allowed measurement assigns bounded complex-linear operators E(S)E(S)E(S) to all outcome subsets, with E(∅)=0E(\varnothing)=0E(∅)=0, E(O)=IE(O)=IE(O)=I, real nonnegative quadratic forms on measurable sets, and weak countable additivity on disjoint measurable sequences, and assigns every u∈Hu\in Hu∈H a finite measure satisfying μu(S)=max⁡{Re⁡⟨u,E(S)u⟩,0}\mu_u(S)=\max\{\operatorname{Re}\langle u,E(S)u\rangle,0\}μu​(S)=max{Re⟨u,E(S)u⟩,0} for measurable SSS; weak additivity requires the corresponding scalar inner-product sums for every pair of vectors. Extended-to-real conversion sends +∞+\infty+∞ to zero. There are no sign, normalization, or integrability hypotheses in this definition: square roots, division, logarithms, the real supremum, and the integral use their total conventions, including e=0\sqrt e=0e​=0 when e<0e<0e<0, 0/0=00/0=00/0=0, and log⁡0=0\log 0=0log0=0; the integration condition remains the strict inequality x0<∣x∣x_0<|x|x0​<∣x∣, even for negative x0x_0x0​.

QAlgorithms.Papers.Independent.Intakethird100.P02.amplifiedLeakage

For arbitrary real numbers n,η,gn,\eta,gn,η,g, this defines the amplified leakage as L(n,η/g)L(n,\eta/g)L(n,η/g), where L(n,s)L(n,s)L(n,s) is the real supremum of the following measurement values. Put a=(1−s)na=\sqrt{(1-s)n}a=(1−s)n​, viewed as a complex number, and let u±∈ℓ2(N;C)u_\pm\in\ell^2(\mathbb N;\mathbb C)u±​∈ℓ2(N;C) be the chosen vectors with coordinates e−∣a∣2/2(±a)j/j!e^{-|a|^2/2}(\pm a)^j/\sqrt{j!}e−∣a∣2/2(±a)j/j!​, or the zero vector if no vector with those coordinates exists. Allow every outcome type OOO with any measurable-space structure and every measurement consisting of bounded complex-linear operators E(S)E(S)E(S), defined for all subsets S⊆OS\subseteq OS⊆O, and finite measures μu\mu_uμu​ for every u∈ℓ2(N;C)u\in\ell^2(\mathbb N;\mathbb C)u∈ℓ2(N;C), subject to E(∅)=0E(\varnothing)=0E(∅)=0, E(O)=IE(O)=IE(O)=I, real nonnegative quadratic forms ⟨u,E(S)u⟩\langle u,E(S)u\rangle⟨u,E(S)u⟩ for measurable SSS, weak countable additivity on disjoint measurable sets, and μu(S)=max⁡{Re⁡⟨u,E(S)u⟩,0}\mu_u(S)=\max\{\operatorname{Re}\langle u,E(S)u\rangle,0\}μu​(S)=max{Re⟨u,E(S)u⟩,0} on measurable sets. Here weak countable additivity means that, for every u,vu,vu,v and disjoint measurable sequence SjS_jSj​, the scalar family ⟨u,E(Sj)v⟩\langle u,E(S_j)v\rangle⟨u,E(Sj​)v⟩ sums to ⟨u,E(⋃jSj)v⟩\langle u,E(\bigcup_jS_j)v\rangle⟨u,E(⋃j​Sj​)v⟩. For each such measurement, set μ±=μu±\mu_\pm=\mu_{u_\pm}μ±​=μu±​​ and p(o)=(dμ+/d(μ++μ−)(o))realp(o)=\bigl(d\mu_+/d(\mu_++\mu_-)(o)\bigr)_{\mathrm{real}}p(o)=(dμ+​/d(μ+​+μ−​)(o))real​, where conversion from an extended nonnegative number to a real sends +∞+\infty+∞ to 000; its measurement value is log⁡2∫O[p(o)2+(1−p(o))2] d(μ++μ−)(o)\log_2\int_O[p(o)^2+(1-p(o))^2]\,d(\mu_++\mu_-)(o)log2​∫O​[p(o)2+(1−p(o))2]d(μ+​+μ−​)(o). Thus the integral uses the sum of the two laws, without a factor 1/21/21/2. No restrictions on n,η,gn,\eta,gn,η,g, or boundedness or attainment of the real supremum, are imposed by this definition; division is total, so g=0g=0g=0 gives η/g=0\eta/g=0η/g=0, and the real square root is zero on negative arguments.

QAlgorithms.Papers.Independent.Intakethird100.P02.amplifiedVariance

For every real ggg, the declared amplified variance is the real number (2g−1)/4(2g-1)/4(2g−1)/4. No nonnegativity hypothesis is imposed: it is zero at g=1/2g=1/2g=1/2 and negative for g<1/2g<1/2g<1/2.

QAlgorithms.Papers.Independent.Intakethird100.P02.binaryDensity

For arbitrary real v,energy,xv,\mathrm{energy},xv,energy,x, let V=max⁡{v,0}V=\max\{v,0\}V=max{v,0} and a=energya=\sqrt{\mathrm{energy}}a=energy​, where a=0a=0a=0 if energy<0\mathrm{energy}<0energy<0. This defines (Ga,V(x)+G−a,V(x))/2\bigl(G_{a,V}(x)+G_{-a,V}(x)\bigr)/2(Ga,V​(x)+G−a,V​(x))/2, where Gμ,VG_{\mu,V}Gμ,V​ is the real Gaussian density function with mean μ\muμ and variance VVV: for V>0V>0V>0,

Gμ,V(x)=12πVexp⁡ ⁣(−(x−μ)22V),G_{\mu,V}(x)=\frac{1}{\sqrt{2\pi V}}\exp\!\left(-\frac{(x-\mu)^2}{2V}\right),Gμ,V​(x)=2πV​1​exp(−2V(x−μ)2​),

and for V=0V=0V=0 this density function is 000. Thus v≤0v\leq0v≤0 makes the defined function identically zero; no positivity assumptions on vvv or energy\mathrm{energy}energy are required.

QAlgorithms.Papers.Independent.Intakethird100.P02.binaryDisplacedLaw

For arbitrary real V,EV,EV,E, this measure is

12 L(E+Re⁡β+t)+12 L(−E+Re⁡β+t),\frac12\,\mathcal L\bigl(\sqrt E+\operatorname{Re}\beta+t\bigr) +\frac12\,\mathcal L\bigl(-\sqrt E+\operatorname{Re}\beta+t\bigr),21​L(E​+Reβ+t)+21​L(−E​+Reβ+t),

where each L\mathcal LL denotes pushforward of the product measure νV⊗γ\nu_V\otimes\gammaνV​⊗γ, γ\gammaγ is the real Gaussian probability measure of mean zero and variance 1/41/41/4, and ν0=δ0\nu_0=\delta_0ν0​=δ0​, while νV\nu_VνV​ for V≠0V\ne0V=0 has planar Lebesgue density max⁡(e−∣β∣2/V/(πV),0)\max(e^{-|\beta|^2/V}/(\pi V),0)max(e−∣β∣2/V/(πV),0). The coefficients 1/21/21/2 are nonnegative extended-real measure scalars. No positivity assumptions are imposed: negative VVV gives the zero measure, and E≤0E\le0E≤0 gives E=0\sqrt E=0E​=0, so the two component measures coincide.

QAlgorithms.Papers.Independent.Intakethird100.P02.binaryInformation

For every real ppp, this defines

1+plog⁡2p+(1−p)log⁡2(1−p).1+p\log_2 p+(1-p)\log_2(1-p).1+plog2​p+(1−p)log2​(1−p).

There is no assumption that 0≤p≤10\leq p\leq10≤p≤1. The real logarithm is total: log⁡20=0\log_2 0=0log2​0=0, and for negative arguments log⁡2x=log⁡∣x∣/log⁡2\log_2 x=\log|x|/\log2log2​x=log∣x∣/log2. In particular, the defined value is 111 at both p=0p=0p=0 and p=1p=1p=1.

QAlgorithms.Papers.Independent.Intakethird100.P02.binaryPosterior

For any measurable space OOO, any two measures μ+,μ−\mu_+,\mu_-μ+​,μ−​ on OOO, and any point o∈Oo\in Oo∈O, this defines the real number p(o)=(dμ+d(μ++μ−)(o))realp(o)=\bigl(\frac{d\mu_+}{d(\mu_++\mu_-)}(o)\bigr)_{\mathrm{real}}p(o)=(d(μ+​+μ−​)dμ+​​(o))real​, using the selected extended-nonnegative-valued Radon–Nikodym derivative and its conversion to a real number. This conversion leaves finite values unchanged and sends +∞+\infty+∞ to 000. No finiteness, probability normalization, or other assumptions on the measures are imposed, and the definition uses the selected derivative at the specified point, including points where its value is not determined by an almost-everywhere characterization.

QAlgorithms.Papers.Independent.Intakethird100.P02.classicalTeleportationHomodyne

For every real η\etaη, complex α\alphaα, and Boolean kkk, this defines the measure on R\mathbb RR obtained by pushing forward να⊗N(0,1/4)\nu_\alpha\otimes\mathcal N(0,1/4)να​⊗N(0,1/4) under

(z,t)⟼η qk(z)+t,dνα(z)=e−∣z−α∣2π d2z,(z,t)\longmapsto \sqrt\eta\,q_k(z)+t, \qquad d\nu_\alpha(z)=\frac{e^{-|z-\alpha|^2}}{\pi}\,d^2z,(z,t)⟼η​qk​(z)+t,dνα​(z)=πe−∣z−α∣2​d2z,

where qfalse(z)=Re⁡zq_{\mathrm{false}}(z)=\operatorname{Re}zqfalse​(z)=Rez and qtrue(z)=Im⁡zq_{\mathrm{true}}(z)=\operatorname{Im}zqtrue​(z)=Imz. Thus the two variables have the product law before applying this map, and 1/41/41/4 is the real Gaussian variance. No restriction on η\etaη is imposed; for η≤0\eta\le0η≤0, its real square root is zero.

QAlgorithms.Papers.Independent.Intakethird100.P02.classicalTeleportationOutput

For every real η\etaη and complex α\alphaα, this defines a complex matrix indexed by all j,k∈Nj,k\in\mathbb Nj,k∈N, including zero, by

Tη,α(j,k)=∫Ccj(η z) ck(η z)‾ e−∣z−α∣2π d2z,cj(w)=e−∣w∣2/2wjj!.T_{\eta,\alpha}(j,k) =\int_{\mathbb C} c_j(\sqrt\eta\,z)\,\overline{c_k(\sqrt\eta\,z)} \,\frac{e^{-|z-\alpha|^2}}{\pi}\,d^2z, \qquad c_j(w)=\frac{e^{-|w|^2/2}w^j}{\sqrt{j!}}.Tη,α​(j,k)=∫C​cj​(η​z)ck​(η​z)​πe−∣z−α∣2​d2z,cj​(w)=j!​e−∣w∣2/2wj​.

The integral is the complex-valued Lebesgue integral against the displayed complex Gaussian measure. No positivity restriction on η\etaη or integrability hypothesis is included in the definition; the real square root satisfies η=0\sqrt\eta=0η​=0 for η≤0\eta\le0η≤0, and the integral uses its total convention when an integrand is not integrable.

QAlgorithms.Papers.Independent.Intakethird100.P02.coherentAmplitudeQuadrature

For every Boolean choice kkk and complex number α\alphaα, the declared quadrature is Re⁡α\operatorname{Re}\alphaReα when kkk is false and Im⁡α\operatorname{Im}\alphaImα when kkk is true.

QAlgorithms.Papers.Independent.Intakethird100.P02.displacedCoherent

For every complex mean mmm and real VVV, this definition gives a complex matrix indexed by all j,k∈Nj,k\in\mathbb Nj,k∈N, including zero, with entries

Dm,V(j,k)=∫Ccj(m+β) ck(m+β)‾ dνV(β),cj(z)=e−∣z∣2/2zjj!.D_{m,V}(j,k)=\int_{\mathbb C}c_j(m+\beta)\, \overline{c_k(m+\beta)}\,d\nu_V(\beta), \qquad c_j(z)=\frac{e^{-|z|^2/2}z^j}{\sqrt{j!}}.Dm,V​(j,k)=∫C​cj​(m+β)ck​(m+β)​dνV​(β),cj​(z)=j!​e−∣z∣2/2zj​.

The measure ν0\nu_0ν0​ is the point mass at zero; for V≠0V\ne0V=0, νV\nu_VνV​ has density max⁡(e−∣β∣2/V/(πV),0)\max(e^{-|\beta|^2/V}/(\pi V),0)max(e−∣β∣2/V/(πV),0) with respect to planar Lebesgue measure. Thus negative VVV gives the zero measure, and V=0V=0V=0 gives Dm,0(j,k)=cj(m)ck(m)‾D_{m,0}(j,k)=c_j(m)\overline{c_k(m)}Dm,0​(j,k)=cj​(m)ck​(m)​. No integrability or density-matrix condition is part of this definition; each entry uses the total Bochner integral, which is zero when its integrand is not integrable.

QAlgorithms.Papers.Independent.Intakethird100.P02.displacedHomodyne

For every complex mmm, real VVV, and Boolean kkk, this definition gives the pushforward of νV⊗γ\nu_V\otimes\gammaνV​⊗γ under

(β,t)⟼Qk(m+β)+t,Qfalse(z)=Re⁡z,Qtrue(z)=Im⁡z,(\beta,t)\longmapsto Q_k(m+\beta)+t, \qquad Q_{\mathrm{false}}(z)=\operatorname{Re}z,\quad Q_{\mathrm{true}}(z)=\operatorname{Im}z,(β,t)⟼Qk​(m+β)+t,Qfalse​(z)=Rez,Qtrue​(z)=Imz,

where γ\gammaγ is the real Gaussian probability measure with mean zero and variance 1/41/41/4, and ν0=δ0\nu_0=\delta_0ν0​=δ0​, while for V≠0V\ne0V=0, νV\nu_VνV​ has planar Lebesgue density max⁡(e−∣β∣2/V/(πV),0)\max(e^{-|\beta|^2/V}/(\pi V),0)max(e−∣β∣2/V/(πV),0). There is no sign restriction on VVV: V=0V=0V=0 gives the real Gaussian measure with mean Qk(m)Q_k(m)Qk​(m) and variance 1/41/41/4, and negative VVV gives the zero measure.

QAlgorithms.Papers.Independent.Intakethird100.P02.gainWithLeakage

For arbitrary real numbers x0,E,v,ℓx_0,E,v,\ellx0​,E,v,ℓ, let fv(x;μ)f_v(x;\mu)fv​(x;μ) be the real Gaussian density with mean μ\muμ and variance max⁡(v,0)\max(v,0)max(v,0), using its total definition, which is zero when that variance is zero. Define

bv,E(x)=fv(x;E)+fv(x;−E)2,pv,E(x)=fv(x;E)fv(x;E)+fv(x;−E),Iv,E(x)=1+pv,E(x)log⁡2pv,E(x)+(1−pv,E(x))log⁡2(1−pv,E(x)).b_{v,E}(x)=\frac{f_v(x;\sqrt E)+f_v(x;-\sqrt E)}2,\qquad p_{v,E}(x)=\frac{f_v(x;\sqrt E)} {f_v(x;\sqrt E)+f_v(x;-\sqrt E)}, \quad I_{v,E}(x)=1+p_{v,E}(x)\log_2p_{v,E}(x) +(1-p_{v,E}(x))\log_2(1-p_{v,E}(x)).bv,E​(x)=2fv​(x;E​)+fv​(x;−E​)​,pv,E​(x)=fv​(x;E​)+fv​(x;−E​)fv​(x;E​)​,Iv,E​(x)=1+pv,E​(x)log2​pv,E​(x)+(1−pv,E​(x))log2​(1−pv,E​(x)).

The declared gain is

G(x0,E,v,ℓ)=12∫{x∈R: x0<∣x∣}bv,E(x)(Iv,E(x)−ℓ) dx.G(x_0,E,v,\ell) =\frac12\int_{\{x\in\mathbb R:\,x_0<|x|\}} b_{v,E}(x)\bigl(I_{v,E}(x)-\ell\bigr)\,dx.G(x0​,E,v,ℓ)=21​∫{x∈R:x0​<∣x∣}​bv,E​(x)(Iv,E​(x)−ℓ)dx.

There are no sign, variance, or integrability hypotheses: real square roots of negative numbers are zero, division by zero is zero, and log⁡0=0\log 0=0log0=0; the integral uses the total Bochner-integral convention and is zero if its integrand is not integrable on the indicated set. The threshold is strict, and a negative x0x_0x0​ makes the integration set all of R\mathbb RR.

QAlgorithms.Papers.Independent.Intakethird100.P02.homodyneInformation

For arbitrary real v,energy,xv,\mathrm{energy},xv,energy,x, set V=max⁡{v,0}V=\max\{v,0\}V=max{v,0}, a=energya=\sqrt{\mathrm{energy}}a=energy​, and

p=Ga,V(x)Ga,V(x)+G−a,V(x).p=\frac{G_{a,V}(x)}{G_{a,V}(x)+G_{-a,V}(x)}.p=Ga,V​(x)+G−a,V​(x)Ga,V​(x)​.

Here a=0a=0a=0 for negative energy, Gμ,V(x)=(2πV)−1/2exp⁡(−(x−μ)2/(2V))G_{\mu,V}(x)=(2\pi V)^{-1/2}\exp(-(x-\mu)^2/(2V))Gμ,V​(x)=(2πV)−1/2exp(−(x−μ)2/(2V)) for V>0V>0V>0, and Gμ,0(x)=0G_{\mu,0}(x)=0Gμ,0​(x)=0; division by zero gives 000. This definition returns 1+plog⁡2p+(1−p)log⁡2(1−p)1+p\log_2p+(1-p)\log_2(1-p)1+plog2​p+(1−p)log2​(1−p), using the total real logarithm with log⁡20=0\log_2 0=0log2​0=0 and log⁡2t=log⁡∣t∣/log⁡2\log_2 t=\log|t|/\log2log2​t=log∣t∣/log2 for nonzero real ttt. In particular, v≤0v\leq0v≤0 gives p=0p=0p=0 and a defined information value of 111.

QAlgorithms.Papers.Independent.Intakethird100.P02.homodynePosterior

For arbitrary real v,energy,xv,\mathrm{energy},xv,energy,x, set V=max⁡{v,0}V=\max\{v,0\}V=max{v,0} and a=energya=\sqrt{\mathrm{energy}}a=energy​, with negative square-root arguments giving 000. This defines

Ga,V(x)Ga,V(x)+G−a,V(x),\frac{G_{a,V}(x)}{G_{a,V}(x)+G_{-a,V}(x)},Ga,V​(x)+G−a,V​(x)Ga,V​(x)​,

where Gμ,V(x)=(2πV)−1/2exp⁡(−(x−μ)2/(2V))G_{\mu,V}(x)=(2\pi V)^{-1/2}\exp(-(x-\mu)^2/(2V))Gμ,V​(x)=(2πV)−1/2exp(−(x−μ)2/(2V)) for V>0V>0V>0, and Gμ,0(x)=0G_{\mu,0}(x)=0Gμ,0​(x)=0. Real division is totalized with division by zero equal to 000; consequently the result is 000 when v≤0v\leq0v≤0. There are no parameter restrictions.

QAlgorithms.Papers.Independent.Intakethird100.P02.leakageValues

For arbitrary real n,ηn,\etan,η, put a=(1−η)n∈Ca=\sqrt{(1-\eta)n}\in\mathbb Ca=(1−η)n​∈C, using the real square root, which is 000 for negative arguments, and let u±=c±a∈H=ℓ2(N;C)u_\pm=c_{\pm a}\in H=\ell^2(\mathbb N;\mathbb C)u±​=c±a​∈H=ℓ2(N;C). Here cαc_\alphacα​ is a classically selected vector whose coordinates are e−∣α∣2/2αj/j!e^{-|\alpha|^2/2}\alpha^j/\sqrt{j!}e−∣α∣2/2αj/j!​ for every j∈Nj\in\mathbb Nj∈N, if such a vector exists in HHH, and is the zero vector otherwise. This defines the set of all real numbers

log⁡2 ⁣(∫O[p(o)2+(1−p(o))2] d(μ++μ−)(o))\log_2\!\left(\int_O[p(o)^2+(1-p(o))^2]\,d(\mu_++\mu_-)(o)\right)log2​(∫O​[p(o)2+(1−p(o))2]d(μ+​+μ−​)(o))

obtained by choosing any type OOO, any measurable-space structure on it, and any measurement with bounded complex-linear effects on HHH and finite laws μu\mu_uμu​. Such effects are assigned to all subsets, take values 000 and IHI_HIH​ on the empty and total sets, have nonnegative real quadratic forms on measurable sets, and are countably additive on disjoint measurable sets after pairing with any two vectors; their laws satisfy μu(S)=max⁡{Re⁡⟨u,E(S)u⟩,0}\mu_u(S)=\max\{\operatorname{Re}\langle u,E(S)u\rangle,0\}μu​(S)=max{Re⟨u,E(S)u⟩,0} on measurable sets. In the displayed expression, μ±=μu±\mu_\pm=\mu_{u_\pm}μ±​=μu±​​ and ppp is the selected Radon–Nikodym derivative of μ+\mu_+μ+​ relative to μ++μ−\mu_++\mu_-μ+​+μ−​, converted to a real number by sending +∞+\infty+∞ to 000. The logarithm uses the total real convention log⁡20=0\log_2 0=0log2​0=0. No parameter restrictions or separate realization or normalization hypotheses on the two selected vectors occur in this definition.

QAlgorithms.Papers.Independent.Intakethird100.P02.measurementRenyi

For any measurable space OOO, any measurement EEE as described here, and arbitrary vectors u+,u−∈H=ℓ2(N;C)u_+,u_-\in H=\ell^2(\mathbb N;\mathbb C)u+​,u−​∈H=ℓ2(N;C), let μ±\mu_\pmμ±​ be their assigned finite laws and put p(o)=(dμ+d(μ++μ−)(o))realp(o)=\bigl(\frac{d\mu_+}{d(\mu_++\mu_-)}(o)\bigr)_{\mathrm{real}}p(o)=(d(μ+​+μ−​)dμ+​​(o))real​, with +∞+\infty+∞ converted to 000. The measurement data consist of bounded complex-linear effects on HHH, assigned to all subsets of OOO, with empty effect 000, total effect IHI_HIH​, nonnegative real quadratic forms on measurable sets, and countable additivity on disjoint measurable sets after pairing with any two vectors; its finite laws satisfy μu(S)=max⁡{Re⁡⟨u,E(S)u⟩,0}\mu_u(S)=\max\{\operatorname{Re}\langle u,E(S)u\rangle,0\}μu​(S)=max{Re⟨u,E(S)u⟩,0} on measurable sets. This definition is

log⁡2 ⁣(∫O[p(o)2+(1−p(o))2] d(μ++μ−)(o)).\log_2\!\left(\int_O \bigl[p(o)^2+(1-p(o))^2\bigr]\,d(\mu_++\mu_-)(o)\right).log2​(∫O​[p(o)2+(1−p(o))2]d(μ+​+μ−​)(o)).

The vectors need not have unit norm, and the sum measure is not divided by its total mass. The logarithm is the total real logarithm to base 222, which takes value 000 at 000; thus the expression is also defined when both vectors are zero.

QAlgorithms.Papers.Independent.Intakethird100.P02.mutualInformationBits

For every pair of types A,OA,OA,O with measurable-space structures and every measure JJJ on A×OA\times OA×O, this defines the extended nonnegative number

DKL ⁣(J ∥ ((pr⁡A)∗J)⊗((pr⁡O)∗J))log⁡2,\frac{D_{\mathrm{KL}}\!\left(J\,\middle\|\, ((\operatorname{pr}_A)_*J)\otimes((\operatorname{pr}_O)_*J)\right)} {\log 2},log2DKL​(J∥((prA​)∗​J)⊗((prO​)∗​J))​,

where the denominator is embedded into the extended nonnegative reals and the two pushforwards are the marginal measures. No probability, finiteness, or absolute-continuity assumption is imposed on JJJ; the value uses the defined extended-valued Kullback–Leibler divergence and may be +∞+\infty+∞.

QAlgorithms.Papers.Independent.Intakethird100.P02.normalizedExcess

For every real vvv, this defines its normalized excess as v/(1/4)−1=4v−1v/(1/4)-1=4v-1v/(1/4)−1=4v−1. There is no nonnegativity or variance assumption on vvv.

QAlgorithms.Papers.Independent.Intakethird100.P02.optimalLeakage

For arbitrary real n,ηn,\etan,η, this defines the real supremum of the following measurement values. Set a=(1−η)n∈Ca=\sqrt{(1-\eta)n}\in\mathbb Ca=(1−η)n​∈C, with negative real square-root arguments giving 000, and let cα∈H=ℓ2(N;C)c_\alpha\in H=\ell^2(\mathbb N;\mathbb C)cα​∈H=ℓ2(N;C) be a selected vector with coordinates e−∣α∣2/2αj/j!e^{-|\alpha|^2/2}\alpha^j/\sqrt{j!}e−∣α∣2/2αj/j!​ if one exists, or 000 otherwise. Vary over every type OOO, every measurable-space structure on OOO, and every assignment of bounded complex-linear effects E(S)E(S)E(S) and finite laws μu\mu_uμu​ satisfying E(∅)=0E(\varnothing)=0E(∅)=0, E(O)=IHE(O)=I_HE(O)=IH​, positivity of quadratic forms on measurable sets, countable additivity on disjoint measurable sets after pairing with arbitrary vectors, and μu(S)=max⁡{Re⁡⟨u,E(S)u⟩,0}\mu_u(S)=\max\{\operatorname{Re}\langle u,E(S)u\rangle,0\}μu​(S)=max{Re⟨u,E(S)u⟩,0} for measurable SSS. For each such choice take

log⁡2 ⁣(∫O[p(o)2+(1−p(o))2] d(μca+μc−a)(o)),p=(dμcad(μca+μc−a))real,\log_2\!\left(\int_O[p(o)^2+(1-p(o))^2]\,d(\mu_{c_a}+\mu_{c_{-a}})(o)\right), \qquad p=\left(\frac{d\mu_{c_a}}{d(\mu_{c_a}+\mu_{c_{-a}})}\right)_{\mathrm{real}},log2​(∫O​[p(o)2+(1−p(o))2]d(μca​​+μc−a​​)(o)),p=(d(μca​​+μc−a​​)dμca​​​)real​,

where conversion sends +∞+\infty+∞ to 000, and the total real logarithm satisfies log⁡20=0\log_2 0=0log2​0=0. The supremum is taken in R\mathbb RR; its totalized value is 000 if the set is empty or unbounded above. The definition itself does not assert that a measurement attains this supremum.

QAlgorithms.Papers.Independent.Intakethird100.P02.reflectedAmplitude

For arbitrary real numbers n,ηn,\etan,η, this defines the complex number a=(1−η)n+0ia=\sqrt{(1-\eta)n}+0ia=(1−η)n​+0i. There are no sign or range restrictions on either parameter. The real square root is totalized to 000 on negative arguments, so a=0a=0a=0 whenever (1−η)n≤0(1-\eta)n\leq0(1−η)n≤0.

QAlgorithms.Papers.Independent.Intakethird100.P02.splittingAmplifierOutput

For arbitrary real η,g\eta,gη,g and complex α\alphaα, this definition gives a complex matrix indexed by pairs of natural numbers. Put a=η αa=\sqrt\eta\,\alphaa=η​α, b=1−η/g αb=\sqrt{1-\eta/g}\,\alphab=1−η/g​α, and cj(z)=e−∣z∣2/2zj/j!c_j(z)=e^{-|z|^2/2}z^j/\sqrt{j!}cj​(z)=e−∣z∣2/2zj/j!​. Its entry at row (j,r)(j,r)(j,r) and column (k,s)(k,s)(k,s), for all j,r,k,s∈Nj,r,k,s\in\mathbb Nj,r,k,s∈N, is

(∫Ccj(a+β)ck(a+β)‾ dνg−1(β))cr(b)cs(b)‾,\left(\int_{\mathbb C}c_j(a+\beta)\overline{c_k(a+\beta)} \,d\nu_{g-1}(\beta)\right)c_r(b)\overline{c_s(b)},(∫C​cj​(a+β)ck​(a+β)​dνg−1​(β))cr​(b)cs​(b)​,

where ν0=δ0\nu_0=\delta_0ν0​=δ0​, and for V≠0V\ne0V=0, νV\nu_VνV​ has planar Lebesgue density max⁡(e−∣β∣2/V/(πV),0)\max(e^{-|\beta|^2/V}/(\pi V),0)max(e−∣β∣2/V/(πV),0). No normalization, positivity, integrability, or parameter restrictions are asserted. In particular, negative square-root arguments give zero square roots, η/g=0\eta/g=0η/g=0 when g=0g=0g=0, g<1g<1g<1 makes the mixing measure zero and hence the whole matrix zero, and g=1g=1g=1 makes the mixing measure a point mass; integrals otherwise use the total Bochner-integral convention.

QAlgorithms.Papers.Independent.Intakethird100.P02.splittingGain

For arbitrary real x0,n,ηx_0,n,\etax0​,n,η, the splitting gain is

S(x0,n,η)=12∫∣x∣>x0b(x)(I(x)−L(n,η)) dx,S(x_0,n,\eta)=\frac12\int_{|x|>x_0} b(x)\bigl(I(x)-L(n,\eta)\bigr)\,dx,S(x0​,n,η)=21​∫∣x∣>x0​​b(x)(I(x)−L(n,η))dx,

where, writing f(x;μ)f(x;\mu)f(x;μ) for the Gaussian density with mean μ\muμ and variance 1/41/41/4,

b(x)=f(x;ηn)+f(x;−ηn)2,p(x)=f(x;ηn)f(x;ηn)+f(x;−ηn),I(x)=1+p(x)log⁡2p(x)+(1−p(x))log⁡2(1−p(x)).b(x)=\frac{f(x;\sqrt{\eta n})+f(x;-\sqrt{\eta n})}{2},\qquad p(x)=\frac{f(x;\sqrt{\eta n})}{f(x;\sqrt{\eta n})+f(x;-\sqrt{\eta n})}, \qquad I(x)=1+p(x)\log_2p(x)+(1-p(x))\log_2(1-p(x)).b(x)=2f(x;ηn​)+f(x;−ηn​)​,p(x)=f(x;ηn​)+f(x;−ηn​)f(x;ηn​)​,I(x)=1+p(x)log2​p(x)+(1−p(x))log2​(1−p(x)).

Here L(n,η)L(n,\eta)L(n,η) is the real supremum, over all measurable outcome spaces and all oscillator measurements described below, of

log⁡2 ⁣(∫(r(o)2+(1−r(o))2) d(μ++μ−)(o)),r(o)=(dμ+d(μ++μ−)(o))R.\log_2\!\left(\int \bigl(r(o)^2+(1-r(o))^2\bigr)\,d(\mu_++\mu_-)(o)\right), \qquad r(o)=\left(\frac{d\mu_+}{d(\mu_++\mu_-)}(o)\right)_{\mathbb R}.log2​(∫(r(o)2+(1−r(o))2)d(μ+​+μ−​)(o)),r(o)=(d(μ+​+μ−​)dμ+​​(o))R​.

The oscillator space is H=ℓ2(N;C)H=\ell^2(\mathbb N;\mathbb C)H=ℓ2(N;C); μ±\mu_\pmμ±​ are the measurement laws of the coherent vectors at amplitudes ±(1−η)n\pm\sqrt{(1-\eta)n}±(1−η)n​, whose prescribed coordinates are e−∣a∣2/2aj/j!e^{-|a|^2/2}a^j/\sqrt{j!}e−∣a∣2/2aj/j!​, with the vector selected if such an HHH-vector exists and zero otherwise. An allowed measurement assigns a bounded complex-linear effect E(A)E(A)E(A) to every outcome set, with E(∅)=0E(\varnothing)=0E(∅)=0, E(O)=1E(O)=1E(O)=1, real nonnegative quadratic forms on measurable sets, and weak countable additivity on disjoint measurable sets; it also supplies a finite measure μu\mu_uμu​ for every u∈Hu\in Hu∈H, satisfying μu(A)=max⁡(Re⁡⟨u,E(A)u⟩,0)\mu_u(A)=\max(\operatorname{Re}\langle u,E(A)u\rangle,0)μu​(A)=max(Re⟨u,E(A)u⟩,0) on measurable sets. The subscript R\mathbb RR denotes conversion from an extended nonnegative real to a real, sending infinity to zero. No restrictions on x0,n,ηx_0,n,\etax0​,n,η, integrability, or attainment of the supremum are imposed; square roots, division, logarithms, integrals, and real suprema use their total definitions.

Human review
  • Endorsed by Shuze Chen · Nov 1, 2026

    Confirmed by the moderator at approval.

  • Endorsed by Gohan · Nov 1, 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