Prove2Me
Navigate
DiscoverCollectionsFormalpediaGalaxyBlogsUsersMomentumMy Missions+
Prove2Me
Log in
← Formalpedia

Definitions for Gaussian relative entropy of entanglement

Definition
QAlgorithms_Papers_Independent_Intakenext100_P46

by Gohan · Oct 31, 2026 · Mathlib 0df444a (Lean v4.33.1)

prove2quantum

The objects the formalized statements of Gaussian relative entropy of entanglement (arXiv:quant-ph/0402109) are stated in. Each was checked against the paper by a blind read-back and a moderator.

  • operatorRelativeEntropy: PDF p.4, (16): extended Umegaki entropy in nats via finite orthogonal compressions of generalized entropy, including its unequal-trace correction. Infinity on support failure is preserved; no covariance expression defines D.
  • densityEntropy: PDF pp.4–5: independently defined spectral entropy in nats. The continuous weight -x log x vanishes at zero; the nonnegative trace retains infinity.
  • ree: PDF p.4: unrestricted REE; no energy, rank or Gaussian restriction.
  • StateBorder: PDF p.4: boundary inside normalized states in trace norm.
  • GroupedPhase: PDF p.2: grouped coordinates, all q before all p.
  • ModeMatrix: PDF p.2: real covariance/exponential arrays, not density operators.
  • groupedCCR: PDF p.2, (3): grouped canonical form with [q_j,p_k]=i delta_jk.
  • IsPhysicalCovariance: PDF p.2: symmetric covariance with alpha-i Delta/2 nonnegative.
  • StrictPhysical: PDF pp.2–3: finite exponential chart, excluding pure normal modes.
  • IsSymplecticMatrix: PDF p.2, (2)–(3): invertible real symplectic transformation.
  • normalDiagonal: PDF p.2: repeat each normal-mode scalar in both blocks.
  • modeEM: PDF p.2: used only with gamma>1/2, never at a pure mode.
  • coth: PDF p.3: scalar coth, used on positive finite normal parameters.
  • CMEM: PDF pp.2–3, (4),(14): CM and finite EM have a common Williamson chart. Existence from a given covariance is a conclusion, not a premise below.
  • HasGaussian: PDF pp.2,4,6: actual Gaussian density with physical symmetric covariance. The accepted Weyl convention uses doubled covariance, hence 2 • alpha. The explicit symmetry guard avoids the defective frozen moments predicate. Trace class and normalization are provided by State.
  • IsGaussianDensity: PDF p.4, (19): all Gaussian states, including displacements and pure modes.
  • gree: PDF p.4, (19), p.12: extended infimum over all actual separable Gaussian states. It also makes sense for non-Gaussian rho; attainment is separate.
  • GaussianBorder: PDF pp.4–6: separability border relative to physical Gaussian covariances. Every positive covariance neighborhood meets inseparable Gaussian states. Means are fixed because displacements are local; pure covariances are included.
  • CenteredGaussianKernel: PDF p.2, (1),(4): centered Gaussianity on a single complete occupation space. This generic interface also covers matrices before a bipartite regrouping.
  • thermalGibbsKernel: PDF p.2, (1),(4): normalized thermal Gibbs density in the number basis. For positive m its diagonal weights are the normalized exp(-m(k+1/2)) weights. The half-quanta cancel against the reciprocal partition function.
  • HasFiniteEM: PDF pp.2,8, (1),(2): actual finite Gibbs exponential, independently of covariance identities. A bounded metaplectic unitary conjugates the normalized thermal Gibbs density; its action is tested on EVERY bounded Weyl operator. Thus the quadratic exponential is specified spectrally on the completed space, without treating unbounded quadratures as bounded matrices or assuming an entropy formula. The source's EM is S^(-T) diag(m,m) S^(-1).
  • GaussianEM: PDF pp.2–3: the finite covariance chart AND its actual Gibbs meaning. The operator bridge is a conclusion of the conversion theorems, never an additional hypothesis on a supplied Gaussian density. CMEM retains the coordinate relation; this predicate adds its independently specified meaning.
  • joinedCoefficients: PDF p.2: regroup the SAME full bipartite density into one occupation space. No marginal, truncation or new state is chosen.
  • Two: PDF pp.3,8: one oscillator per party, retaining the named A|B cut.
  • pairValues: PDF p.3: two mode values in party order.
  • symmetricModeBlock: PDF pp.3,8: symmetric real two-by-two block [[a,c],[c,b]].
  • standardCM: PDF p.8, (36): ordinary covariance convention, without a factor 1/2.
  • symmetricCM: PDF p.3, (12): half-covariance convention specific to m,kq,kp.
  • symmetricSq: PDF p.3, (13): the symmetric-state block Sq.
  • symplecticLift: PDF p.3: a block symplectic candidate; invertibility is proved in the results.
  • symmetricGammas: PDF p.3, following (14): symplectic eigenvalues, including the pure endpoint.
  • emFrom: PDF p.3, (14): the finite EM constructed from a specified normal chart.
  • normalization: PDF p.4, (17): reciprocal partition function, not the partition function.
  • bosonic: PDF p.5: bosonic entropy in nats, with 0 log 0=0.
  • covarianceObjective: PDF p.4, (18): the finite covariance expression, not the definition of D.
  • displacementPenalty: PDF p.6, (28): the displacement correction with z in the source Weyl convention W(z)=exp(i F^T z). The displaced density mean is -Delta z.
  • weylDisplacedDensity: PDF p.6, (27): the actual W(z) sigma W(z)^* coefficients. The accepted Weyl argument is xi=Delta*z, since W_xi=exp(i xi^T Delta F)=exp(i F^T z).
  • hyperbolicMixing: PDF pp.5,8: two-mode squeezing R(r).
  • modeRotation: PDF p.5, (25): passive rotation, with the source's signs.
  • xSqueeze: PDF pp.7–9: the actual symplectic X(x), with inverse p scalings.
  • squeezeBoth: PDF p.8: a two-mode squeeze acting with opposite r on p.
  • rotateBoth: PDF p.8: the same passive rotation on q and p.
  • ySqueeze: PDF p.7: equal local squeezes Y(y), with y>0 at every use.
  • congruence: PDF p.8: covariance congruence, with its transpose explicit.
  • covarianceType: PDF p.8: the symmetric chart takes precedence over overlapping raw ratios.
  • CovSeparable: PDF pp.8–9: physical Gaussian covariance separability on the actual 1|1 cut.
  • CovBorder: PDF pp.8–9: physical border adjacent to entangled Gaussian states.
  • typeICM: PDF p.8: reverse the type-I diagonalization to construct the covariance.
  • typeIICM: PDF pp.8–9: reverse the type-II diagonalization.
  • TypeIIDomain: PDF p.9, following (38): nontrivial type-II parameter chart.
  • borderLeft: PDF pp.8–9, (37),(38): common left-hand side, with unsquared factor 1/2.
  • borderRightI: PDF p.8, (37): plus sign on the squared-eigenvalue sum.
  • borderRightII: PDF p.9, (38): minus sign on the squared-eigenvalue sum.
  • PositiveRoots: PDF p.9: solve x^2+x^(-2)=K in its positive domain.
  • exceptionalDelta: PDF p.9: delta=(gamma_A^2-1/4)(gamma_B^2-1/4), not a symplectic form.
  • exceptionalLower: PDF p.9: first triangular type-III border Sq.
  • exceptionalUpper: PDF p.9: second triangular type-III border Sq.
  • symmetricBorderQ: PDF p.9, (39): symmetric border Sq in the p.3 normalization.
  • borderCMFrom: PDF p.9: construct a covariance from the displayed Sq and symplectic spectrum.
  • symmetricCandidateCM: PDF pp.9,11, (39),(41): a symmetric border candidate, after the remaining common local squeezing. This is a covariance construction even at pure modes.
  • symmetricValue: PDF p.11, (41): two-variable operator-defined objective, including endpoints. For fixed normal covariances, eliminate y>0 and evaluate actual Umegaki entropy of the explicitly constructed density. Uniqueness of Gaussian characteristic functions means the state variable introduces no extra variational freedom. At gamma=1/2 this definition uses the actual singular density, not Real.log 0, not an assigned scalar limit, and not GREE itself.
  • symmetricObjective: PDF p.11, (41): finite scalar chart of the independent operator objective. The factor 1/2 multiplies the entire square root; coth and tanh each occur as products at half arguments. The state entropy is added separately.
  • tmstObjective: PDF p.11, (42): one-variable objective, with u the positive finite EM normal parameter. This is not used to define the operator divergence or GREE.
  • tmstBorderCM: PDF pp.9,11: a symmetric TMST border with equal normal eigenvalues.
Definition code
import Mathlib
import Definitions.Def_QAlgorithms_Papers_Independent_Dependency50_P13
import Definitions.Def_QAlgorithms_Papers_Independent_Dependency50_P33
import Definitions.Def_QAlgorithms_Papers_Independent_Claimed50_P47
import Definitions.Def_QAlgorithms_Papers_Independent_Next46_P13
import Definitions.Def_QAlgorithms_Papers_Independent_Frontier18_P18
import Definitions.Def_QAlgorithms_Papers_Independent_Claimed50_P41

/-!
Chen, arXiv:quant-ph/0402109v3; PDF pages throughout.
Accepted-paper imports follow the independent-paper override. Full oscillator
states have no cutoff. Divergence is independent of covariance formulas.
Covariances are symmetric with vacuum I/2. Pure modes remain in the state space;
CMEM is only a finite chart. HasBipartiteMoments is never used.
Compilation is pending the separate mechanical handoff.
-/

noncomputable section

namespace QAlgorithms.Papers.Independent.Intakenext100.P46

open scoped BigOperators ENNReal ComplexOrder Matrix

attribute [local instance] Classical.propDecidable Classical.decEq

open QAlgorithms.RealRegisters (Hilbert)
open QAlgorithms.Papers.Independent.Dependency50.P33 (State IsSeparable)
open QAlgorithms.Papers.Independent.Dependency50.P13
  (finiteRelativeEntropy compression entropyWeight)
open QAlgorithms.Papers.Independent.Claimed50.P47
  (Occupation bipartiteCharacteristic bipartiteWeyl gaussianStateCharacteristic)

-- These accepted definitions have the same completed Hilbert-space meaning.
open QAlgorithms.Papers.Independent.Next46.P13 (coordinateBasis)
open QAlgorithms.Papers.Independent.Frontier18.P18 (operatorCoeff boundedOperator)
open QAlgorithms.Papers.Independent.Claimed50.P41 (IsDiagonal)

/-- PDF p.4, (16): extended Umegaki entropy in nats via finite orthogonal
compressions of generalized entropy, including its unequal-trace correction.
Infinity on support failure is preserved; no covariance expression defines D. -/
def operatorRelativeEntropy {I : Type} (rho sigma : State I) : ℝ≥0∞ :=
  ⨆ n : ℕ, ⨆ v : {v : Fin n → Hilbert I // Orthonormal ℂ v},
    finiteRelativeEntropy (compression v.val (boundedOperator rho.coefficients))
      (compression v.val (boundedOperator sigma.coefficients))

/-- PDF pp.4–5: independently defined spectral entropy in nats. The continuous
weight -x log x vanishes at zero; the nonnegative trace retains infinity. -/
def densityEntropy {I : Type} (rho : State I) : ℝ≥0∞ := by
  letI : ContinuousFunctionalCalculus ℂ (Hilbert I →L[ℂ] Hilbert I) IsStarNormal :=
    IsStarNormal.instContinuousFunctionalCalculus (A := Hilbert I →L[ℂ] Hilbert I)
  letI : ContinuousFunctionalCalculus ℝ (Hilbert I →L[ℂ] Hilbert I) IsSelfAdjoint :=
    IsSelfAdjoint.instContinuousFunctionalCalculus (A := Hilbert I →L[ℂ] Hilbert I)
  exact ∑' i, ENNReal.ofReal
    (inner ℂ (coordinateBasis I i)
      ((cfc entropyWeight (boundedOperator rho.coefficients)) (coordinateBasis I i))).re

/-- PDF p.4: unrestricted REE; no energy, rank or Gaussian restriction. -/
def ree {I J : Type} (rho : State (I × J)) : ℝ≥0∞ :=
  ⨅ sigma : {sigma : State (I × J) // IsSeparable sigma.coefficients},
    operatorRelativeEntropy rho sigma.val

/-- PDF p.4: boundary inside normalized states in trace norm. -/
def StateBorder {I J : Type} (sigma : State (I × J)) : Prop :=
  sigma ∈ frontier {tau : State (I × J) | IsSeparable tau.coefficients}

/-- PDF p.2: grouped coordinates, all q before all p. -/
abbrev GroupedPhase (I : Type) := I ⊕ I

/-- PDF p.2: real covariance/exponential arrays, not density operators. -/
abbrev ModeMatrix (I : Type) := Matrix (GroupedPhase I) (GroupedPhase I) ℝ

/-- PDF p.2, (3): grouped canonical form with [q_j,p_k]=i delta_jk. -/
def groupedCCR (I : Type) [Fintype I] : ModeMatrix I := Matrix.fromBlocks 0 1 (-1) 0

/-- PDF p.2: symmetric covariance with alpha-i Delta/2 nonnegative. -/
def IsPhysicalCovariance {I : Type} [Fintype I] (alpha : ModeMatrix I) : Prop :=
  alpha.transpose = alpha ∧
    (alpha.map (fun x : ℝ => (x : ℂ)) -
      (Complex.I / 2) • (groupedCCR I).map (fun x : ℝ => (x : ℂ))).PosSemidef

/-- PDF pp.2–3: finite exponential chart, excluding pure normal modes. -/
def StrictPhysical {I : Type} [Fintype I] (alpha : ModeMatrix I) : Prop :=
  alpha.transpose = alpha ∧
    (alpha.map (fun x : ℝ => (x : ℂ)) -
      (Complex.I / 2) • (groupedCCR I).map (fun x : ℝ => (x : ℂ))).PosDef

/-- PDF p.2, (2)–(3): invertible real symplectic transformation. -/
def IsSymplecticMatrix {I : Type} [Fintype I] (S : ModeMatrix I) : Prop :=
  IsUnit S ∧ S * groupedCCR I * S.transpose = groupedCCR I

/-- PDF p.2: repeat each normal-mode scalar in both blocks. -/
def normalDiagonal {I : Type} [Fintype I] (g : I → ℝ) : ModeMatrix I :=
  Matrix.fromBlocks (Matrix.diagonal g) 0 0 (Matrix.diagonal g)

/-- PDF p.2: used only with gamma>1/2, never at a pure mode. -/
def modeEM (gamma : ℝ) : ℝ := Real.log ((2 * gamma + 1) / (2 * gamma - 1))

/-- PDF p.3: scalar coth, used on positive finite normal parameters. -/
def coth (x : ℝ) : ℝ := Real.cosh x / Real.sinh x

/-- PDF pp.2–3, (4),(14): CM and finite EM have a common Williamson chart.
Existence from a given covariance is a conclusion, not a premise below. -/
def CMEM {I : Type} [Fintype I] (alpha M : ModeMatrix I) : Prop :=
  ∃ (S : ModeMatrix I) (g : I → ℝ), IsSymplecticMatrix S ∧ (∀ j, 1 / 2 < g j) ∧
    alpha = S * normalDiagonal g * S.transpose ∧
    M = S.transpose⁻¹ * normalDiagonal (modeEM ∘ g) * S⁻¹

/-- PDF pp.2,4,6: actual Gaussian density with physical symmetric covariance.
The accepted Weyl convention uses doubled covariance, hence 2 • alpha.
The explicit symmetry guard avoids the defective frozen moments predicate.
Trace class and normalization are provided by State. -/
def HasGaussian {A B : Type} [Fintype A] [Fintype B]
    (rho : State (Occupation A × Occupation B))
    (d : GroupedPhase (A ⊕ B) → ℝ) (alpha : ModeMatrix (A ⊕ B)) : Prop :=
  IsPhysicalCovariance alpha ∧ ∀ xi,
    (∀ u, Summable (fun v => rho.coefficients u v * bipartiteWeyl xi v u)) ∧
    Summable (fun u => ∑' v, rho.coefficients u v * bipartiteWeyl xi v u) ∧
    bipartiteCharacteristic rho.coefficients xi =
      gaussianStateCharacteristic d (2 • alpha) xi

/-- PDF p.4, (19): all Gaussian states, including displacements and pure modes. -/
def IsGaussianDensity {A B : Type} [Fintype A] [Fintype B]
    (rho : State (Occupation A × Occupation B)) : Prop :=
  ∃ d alpha, HasGaussian rho d alpha

/-- PDF p.4, (19), p.12: extended infimum over all actual separable Gaussian
states. It also makes sense for non-Gaussian rho; attainment is separate. -/
def gree {A B : Type} [Fintype A] [Fintype B]
    (rho : State (Occupation A × Occupation B)) : ℝ≥0∞ :=
  ⨅ sigma : {sigma : State (Occupation A × Occupation B) //
      IsSeparable sigma.coefficients ∧ IsGaussianDensity sigma}, operatorRelativeEntropy rho sigma.val

/-- PDF pp.4–6: separability border relative to physical Gaussian covariances.
Every positive covariance neighborhood meets inseparable Gaussian states.
Means are fixed because displacements are local; pure covariances are included. -/
def GaussianBorder {A B : Type} [Fintype A] [Fintype B]
    (sigma : State (Occupation A × Occupation B)) : Prop :=
  IsSeparable sigma.coefficients ∧ ∃ d alpha, HasGaussian sigma d alpha ∧
    ∀ epsilon : ℝ, 0 < epsilon →
      ∃ (tau : State (Occupation A × Occupation B)) (beta : ModeMatrix (A ⊕ B)),
        HasGaussian tau d beta ∧ ¬ IsSeparable tau.coefficients ∧
        ∀ i j, |beta i j - alpha i j| < epsilon

/-- PDF p.2, (1),(4): centered Gaussianity on a single complete occupation space.
This generic interface also covers matrices before a bipartite regrouping. -/
def CenteredGaussianKernel {I : Type} [Fintype I]
    (rho : Matrix (Occupation I) (Occupation I) ℂ) (alpha : ModeMatrix I) : Prop :=
  QAlgorithms.Papers.Independent.Dependency50.P33.IsDensity rho ∧ IsPhysicalCovariance alpha ∧
    ∀ xi,
      (∀ u, Summable (fun v => rho u v *
        QAlgorithms.Papers.Independent.Claimed50.P47.weylKernel xi v u)) ∧
      Summable (fun u => ∑' v, rho u v *
        QAlgorithms.Papers.Independent.Claimed50.P47.weylKernel xi v u) ∧
      QAlgorithms.Papers.Independent.Claimed50.P47.singleCharacteristic rho xi =
        gaussianStateCharacteristic 0 (2 • alpha) xi

/-- PDF p.2, (1),(4): normalized thermal Gibbs density in the number basis.
For positive m its diagonal weights are the normalized exp(-m(k+1/2)) weights.
The half-quanta cancel against the reciprocal partition function. -/
def thermalGibbsKernel {I : Type} [Fintype I] (m : I → ℝ) :
    Matrix (Occupation I) (Occupation I) ℂ :=
  fun u v => if u = v then
    ((∏ j, (1 - Real.exp (-m j)) * Real.exp (-m j * (u j : ℝ)) : ℝ) : ℂ)
    else 0

/-- PDF pp.2,8, (1),(2): actual finite Gibbs exponential, independently of
covariance identities. A bounded metaplectic unitary conjugates the normalized
thermal Gibbs density; its action is tested on EVERY bounded Weyl operator.
Thus the quadratic exponential is specified spectrally on the completed space,
without treating unbounded quadratures as bounded matrices or assuming an
entropy formula. The source's EM is S^(-T) diag(m,m) S^(-1). -/
def HasFiniteEM {I : Type} [Fintype I]
    (rho : Matrix (Occupation I) (Occupation I) ℂ) (M : ModeMatrix I) : Prop :=
  QAlgorithms.Papers.Independent.Dependency50.P33.IsDensity rho ∧
    ∃ (S : ModeMatrix I) (m : I → ℝ)
      (U : Hilbert (Occupation I) ≃ₗᵢ[ℂ] Hilbert (Occupation I)),
      IsSymplecticMatrix S ∧ (∀ j, 0 < m j) ∧
      (∀ xi,
        QAlgorithms.Papers.Independent.Frontier30.P14.boundedUnitaryConjugation U.symm
          (QAlgorithms.Papers.Independent.Claimed50.P47.weylKernel xi) =
        QAlgorithms.Papers.Independent.Claimed50.P47.weylKernel
          (QAlgorithms.Papers.Independent.Claimed50.P47.weylPullback S xi)) ∧
      rho = QAlgorithms.Papers.Independent.Frontier30.P14.boundedUnitaryConjugation U
        (thermalGibbsKernel m) ∧
      M = S.transpose⁻¹ * normalDiagonal m * S⁻¹

/-- PDF pp.2–3: the finite covariance chart AND its actual Gibbs meaning.
The operator bridge is a conclusion of the conversion theorems, never an
additional hypothesis on a supplied Gaussian density. CMEM retains the
coordinate relation; this predicate adds its independently specified meaning. -/
def GaussianEM {I : Type} [Fintype I] (alpha M : ModeMatrix I) : Prop :=
  CMEM alpha M ∧ ∀ rho : Matrix (Occupation I) (Occupation I) ℂ,
    CenteredGaussianKernel rho alpha → HasFiniteEM rho M

/-- PDF p.2: regroup the SAME full bipartite density into one occupation space.
No marginal, truncation or new state is chosen. -/
def joinedCoefficients {A B : Type}
    (rho : Matrix (Occupation A × Occupation B) (Occupation A × Occupation B) ℂ) :
    Matrix (Occupation (A ⊕ B)) (Occupation (A ⊕ B)) ℂ :=
  fun u v => rho (u ∘ Sum.inl, u ∘ Sum.inr) (v ∘ Sum.inl, v ∘ Sum.inr)

example {I : Type} [Fintype I] (m : I → ℝ) (u v : Occupation I) (h : u ≠ v) :
    thermalGibbsKernel m u v = 0 := by simp [thermalGibbsKernel, h]

example {I : Type} [Fintype I] (m : I → ℝ) :
    thermalGibbsKernel m (fun _ => 0) (fun _ => 0) =
      ((∏ j, (1 - Real.exp (-m j)) : ℝ) : ℂ) := by
  simp [thermalGibbsKernel]

example {I : Type} : operatorCoeff (0 : Hilbert I →L[ℂ] Hilbert I) = 0 := by
  ext i j
  simp [operatorCoeff]

example : entropyWeight 0 = 0 := by simp [entropyWeight]

example {I : Type} [Fintype I] (g : I → ℝ) (i : I) :
    normalDiagonal g (Sum.inl i) (Sum.inr i) = 0 := by
  simp [normalDiagonal]

example {I : Type} [Fintype I] (i : I) : groupedCCR I (Sum.inl i) (Sum.inr i) = 1 := by
  simp [groupedCCR]

/-- PDF pp.3,8: one oscillator per party, retaining the named A|B cut. -/
abbrev Two := Fin 1 ⊕ Fin 1

/-- PDF p.3: two mode values in party order. -/
def pairValues (a b : ℝ) : Two → ℝ := Sum.elim (fun _ => a) (fun _ => b)

/-- PDF pp.3,8: symmetric real two-by-two block [[a,c],[c,b]]. -/
def symmetricModeBlock (a c b : ℝ) : Matrix Two Two ℝ
  | Sum.inl _, Sum.inl _ => a
  | Sum.inl _, Sum.inr _ => c
  | Sum.inr _, Sum.inl _ => c
  | Sum.inr _, Sum.inr _ => b

/-- PDF p.8, (36): ordinary covariance convention, without a factor 1/2. -/
def standardCM (a b c1 c2 : ℝ) : ModeMatrix Two :=
  Matrix.fromBlocks (symmetricModeBlock a c1 b) 0 0 (symmetricModeBlock a (-c2) b)

/-- PDF p.3, (12): half-covariance convention specific to m,kq,kp. -/
def symmetricCM (m kq kp : ℝ) : ModeMatrix Two :=
  standardCM (m / 2) (m / 2) (kq / 2) (kp / 2)

/-- PDF p.3, (13): the symmetric-state block Sq. -/
def symmetricSq (m kq kp : ℝ) : Matrix Two Two ℝ :=
  (Real.sqrt 2)⁻¹ • (symmetricModeBlock 1 1 (-1) * Matrix.diagonal
    (pairValues (Real.rpow ((m + kq) / (m - kp)) (1 / 4))
      (Real.rpow ((m - kq) / (m + kp)) (1 / 4))))

/-- PDF p.3: a block symplectic candidate; invertibility is proved in the results. -/
def symplecticLift (Sq : Matrix Two Two ℝ) : ModeMatrix Two :=
  Matrix.fromBlocks Sq 0 0 Sq.transpose⁻¹

/-- PDF p.3, following (14): symplectic eigenvalues, including the pure endpoint. -/
def symmetricGammas (m kq kp : ℝ) : Two → ℝ :=
  pairValues (Real.sqrt ((m + kq) * (m - kp)) / 2)
    (Real.sqrt ((m - kq) * (m + kp)) / 2)

/-- PDF p.3, (14): the finite EM constructed from a specified normal chart. -/
def emFrom {I : Type} [Fintype I] (S : ModeMatrix I) (g : I → ℝ) : ModeMatrix I :=
  S.transpose⁻¹ * normalDiagonal (modeEM ∘ g) * S⁻¹

example : symmetricModeBlock 1 0 1 = (1 : Matrix Two Two ℝ) := by
  ext i j
  rcases i with i | i <;> rcases j with j | j <;>
    fin_cases i <;> fin_cases j <;> simp [symmetricModeBlock]

example (m kq kp : ℝ) :
    symmetricCM m kq kp (Sum.inl (Sum.inl 0)) (Sum.inl (Sum.inr 0)) = kq / 2 := rfl

/-- PDF p.4, (17): reciprocal partition function, not the partition function. -/
def normalization {I : Type} [Fintype I] (g : I → ℝ) : ℝ :=
  ∏ j, 2 * Real.sinh (modeEM (g j) / 2)

/-- PDF p.5: bosonic entropy in nats, with 0 log 0=0. -/
def bosonic (x : ℝ) : ℝ := (x + 1) * Real.log (x + 1) - x * Real.log x

/-- PDF p.4, (18): the finite covariance expression, not the definition of D. -/
def covarianceObjective {I : Type} [Fintype I]
    (alpha M : ModeMatrix I) (c : ℝ) : ℝ := -Real.log c + (alpha * M).trace / 2

example : bosonic 0 = 0 := by simp [bosonic]

/-- PDF p.6, (28): the displacement correction with z in the source Weyl
convention W(z)=exp(i F^T z). The displaced density mean is -Delta z. -/
def displacementPenalty {I : Type} [Fintype I]
    (M : ModeMatrix I) (z : GroupedPhase I → ℝ) : ℝ :=
  (∑ i, ∑ j, (groupedCCR I).mulVec z i * M i j * (groupedCCR I).mulVec z j) / 2

example {I : Type} [Fintype I] (M : ModeMatrix I) : displacementPenalty M 0 = 0 := by
  simp [displacementPenalty]

/-- PDF p.6, (27): the actual W(z) sigma W(z)^* coefficients. The accepted
Weyl argument is xi=Delta*z, since W_xi=exp(i xi^T Delta F)=exp(i F^T z). -/
def weylDisplacedDensity {A B : Type} [Fintype A] [Fintype B]
    (sigma : State (Occupation A × Occupation B)) (z : GroupedPhase (A ⊕ B) → ℝ) :
    Matrix (Occupation A × Occupation B) (Occupation A × Occupation B) ℂ :=
  QAlgorithms.Papers.Independent.Frontier30.P14.conjugateKernel
    (bipartiteWeyl ((groupedCCR (A ⊕ B)).mulVec z)) sigma.coefficients

/-- PDF pp.5,8: two-mode squeezing R(r). -/
def hyperbolicMixing (r : ℝ) : Matrix Two Two ℝ :=
  symmetricModeBlock (Real.cosh r) (Real.sinh r) (Real.cosh r)

/-- PDF p.5, (25): passive rotation, with the source's signs. -/
def modeRotation (t : ℝ) : Matrix Two Two ℝ
  | Sum.inl _, Sum.inl _ => Real.cos t
  | Sum.inl _, Sum.inr _ => Real.sin t
  | Sum.inr _, Sum.inl _ => -Real.sin t
  | Sum.inr _, Sum.inr _ => Real.cos t

/-- PDF pp.7–9: the actual symplectic X(x), with inverse p scalings. -/
def xSqueeze (x : ℝ) : ModeMatrix Two :=
  Matrix.fromBlocks (Matrix.diagonal (pairValues (Real.sqrt x) ((Real.sqrt x)⁻¹))) 0 0
    (Matrix.diagonal (pairValues ((Real.sqrt x)⁻¹) (Real.sqrt x)))

/-- PDF p.8: a two-mode squeeze acting with opposite r on p. -/
def squeezeBoth (r : ℝ) : ModeMatrix Two := Matrix.fromBlocks (hyperbolicMixing r) 0 0 (hyperbolicMixing (-r))

/-- PDF p.8: the same passive rotation on q and p. -/
def rotateBoth (t : ℝ) : ModeMatrix Two := Matrix.fromBlocks (modeRotation t) 0 0 (modeRotation t)

/-- PDF p.7: equal local squeezes Y(y), with y>0 at every use. -/
def ySqueeze (y : ℝ) : ModeMatrix Two :=
  Matrix.fromBlocks ((Real.sqrt y) • (1 : Matrix Two Two ℝ)) 0 0
    ((Real.sqrt y)⁻¹ • (1 : Matrix Two Two ℝ))

/-- PDF p.8: covariance congruence, with its transpose explicit. -/
def congruence (S alpha : ModeMatrix Two) : ModeMatrix Two := S * alpha * S.transpose

/-- PDF p.8: the symmetric chart takes precedence over overlapping raw ratios. -/
def covarianceType (a b c1 c2 : ℝ) : ℕ :=
  if a = b then 4 else if c1 / c2 + c2 / c1 < a / b + b / a then 1
  else if a / b + b / a < c1 / c2 + c2 / c1 then 2 else 3

example (a c1 c2 : ℝ) : covarianceType a a c1 c2 = 4 := by
  simp [covarianceType]

example : modeRotation 0 = (1 : Matrix Two Two ℝ) := by
  ext i j
  rcases i with i | i <;> rcases j with j | j <;>
    fin_cases i <;> fin_cases j <;> simp [modeRotation]

/-- PDF pp.8–9: physical Gaussian covariance separability on the actual 1|1 cut. -/
def CovSeparable (alpha : ModeMatrix Two) : Prop :=
  ∃ sigma : State (Occupation (Fin 1) × Occupation (Fin 1)),
    HasGaussian sigma 0 alpha ∧ IsSeparable sigma.coefficients

/-- PDF pp.8–9: physical border adjacent to entangled Gaussian states. -/
def CovBorder (alpha : ModeMatrix Two) : Prop :=
  ∃ sigma : State (Occupation (Fin 1) × Occupation (Fin 1)),
    HasGaussian sigma 0 alpha ∧ GaussianBorder sigma

/-- PDF p.8: reverse the type-I diagonalization to construct the covariance. -/
def typeICM (gA gB r x : ℝ) : ModeMatrix Two :=
  congruence (squeezeBoth (-r) * xSqueeze x⁻¹) (normalDiagonal (pairValues gA gB))

/-- PDF pp.8–9: reverse the type-II diagonalization. -/
def typeIICM (gA gB t x : ℝ) : ModeMatrix Two :=
  congruence (rotateBoth (-t) * xSqueeze x⁻¹) (normalDiagonal (pairValues gA gB))

/-- PDF p.9, following (38): nontrivial type-II parameter chart. -/
def TypeIIDomain (gA gB x : ℝ) : Prop :=
  0 < x ∧ (max (Real.sqrt (gA / gB)) (Real.sqrt (gB / gA)) < x ∨
    x < min (Real.sqrt (gA / gB)) (Real.sqrt (gB / gA)))

/-- PDF pp.8–9, (37),(38): common left-hand side, with unsquared factor 1/2. -/
def borderLeft (gA gB : ℝ) : ℝ := (2 * gA ^ 2 - 1 / 2) * (2 * gB ^ 2 - 1 / 2)

/-- PDF p.8, (37): plus sign on the squared-eigenvalue sum. -/
def borderRightI (gA gB r x : ℝ) : ℝ :=
  Real.sinh (2 * r) ^ 2 * ((x ^ 2 + (x⁻¹) ^ 2) * gA * gB + (gA ^ 2 + gB ^ 2))

/-- PDF p.9, (38): minus sign on the squared-eigenvalue sum. -/
def borderRightII (gA gB t x : ℝ) : ℝ :=
  Real.sin (2 * t) ^ 2 * ((x ^ 2 + (x⁻¹) ^ 2) * gA * gB - (gA ^ 2 + gB ^ 2))

/-- PDF p.9: solve x^2+x^(-2)=K in its positive domain. -/
def PositiveRoots (K x : ℝ) : Prop :=
  2 ≤ K ∧ (x = Real.sqrt ((K + Real.sqrt (K ^ 2 - 4)) / 2) ∨
    x = Real.sqrt ((K - Real.sqrt (K ^ 2 - 4)) / 2))

example (gA gB x : ℝ) : borderRightI gA gB 0 x = 0 := by
  simp [borderRightI]

/-- PDF p.9: delta=(gamma_A^2-1/4)(gamma_B^2-1/4), not a symplectic form. -/
def exceptionalDelta (gA gB : ℝ) : ℝ := (gA ^ 2 - 1 / 4) * (gB ^ 2 - 1 / 4)

/-- PDF p.9: first triangular type-III border Sq. -/
def exceptionalLower (gA gB : ℝ) : Matrix Two Two ℝ :=
  let d := exceptionalDelta gA gB
  fun i j => match i, j with
    | Sum.inl _, Sum.inl _ => Real.rpow (1 + d / gA ^ 2) (1 / 4)
    | Sum.inl _, Sum.inr _ => 0
    | Sum.inr _, Sum.inl _ => Real.rpow (d ^ 2 / (gA ^ 2 * (gB ^ 2 + d))) (1 / 4)
    | Sum.inr _, Sum.inr _ => Real.rpow (gB ^ 2 / (gB ^ 2 + d)) (1 / 4)

/-- PDF p.9: second triangular type-III border Sq. -/
def exceptionalUpper (gA gB : ℝ) : Matrix Two Two ℝ :=
  let d := exceptionalDelta gA gB
  fun i j => match i, j with
    | Sum.inl _, Sum.inl _ => Real.rpow (gA ^ 2 / (gA ^ 2 + d)) (1 / 4)
    | Sum.inl _, Sum.inr _ => Real.rpow (d ^ 2 / (gB ^ 2 * (gA ^ 2 + d))) (1 / 4)
    | Sum.inr _, Sum.inl _ => 0
    | Sum.inr _, Sum.inr _ => Real.rpow (1 + d / gB ^ 2) (1 / 4)

/-- PDF p.9, (39): symmetric border Sq in the p.3 normalization. -/
def symmetricBorderQ (gA gB : ℝ) : Matrix Two Two ℝ :=
  (Real.sqrt 2)⁻¹ • (symmetricModeBlock 1 1 (-1) * Matrix.diagonal (pairValues
    (Real.rpow (4 * gA ^ 2 * (4 * gB ^ 2 + 1) / (4 * gA ^ 2 + 1)) (1 / 4))
    (Real.rpow ((4 * gB ^ 2 + 1) / (4 * gB ^ 2 * (4 * gA ^ 2 + 1))) (1 / 4))))

/-- PDF p.9: construct a covariance from the displayed Sq and symplectic spectrum. -/
def borderCMFrom (Sq : Matrix Two Two ℝ) (gA gB : ℝ) : ModeMatrix Two :=
  congruence (symplecticLift Sq) (normalDiagonal (pairValues gA gB))

example : exceptionalDelta (1 / 2) (1 / 2) = 0 := by
  norm_num [exceptionalDelta]

/-- PDF pp.9,11, (39),(41): a symmetric border candidate, after the remaining
common local squeezing. This is a covariance construction even at pure modes. -/
def symmetricCandidateCM (gA gB y : ℝ) : ModeMatrix Two :=
  congruence (ySqueeze y⁻¹) (borderCMFrom (symmetricBorderQ gA gB) gA gB)

/-- PDF p.11, (41): two-variable operator-defined objective, including endpoints.
For fixed normal covariances, eliminate y>0 and evaluate actual Umegaki entropy
of the explicitly constructed density. Uniqueness of Gaussian characteristic
functions means the state variable introduces no extra variational freedom.
At gamma=1/2 this definition uses the actual singular density, not Real.log 0,
not an assigned scalar limit, and not GREE itself. -/
def symmetricValue
    (rho : State (Occupation (Fin 1) × Occupation (Fin 1))) (gA gB : ℝ) : ℝ≥0∞ :=
  ⨅ candidate : {p : ℝ × State (Occupation (Fin 1) × Occupation (Fin 1)) //
      0 < p.1 ∧ HasGaussian p.2 0 (symmetricCandidateCM gA gB p.1)},
    operatorRelativeEntropy rho candidate.val.2

/-- PDF p.11, (41): finite scalar chart of the independent operator objective.
The factor 1/2 multiplies the entire square root; coth and tanh each occur as
products at half arguments. The state entropy is added separately. -/
def symmetricObjective (m kq kp u v : ℝ) : ℝ :=
  -Real.log (2 * Real.sinh (u / 2)) - Real.log (2 * Real.sinh (v / 2)) +
    Real.sqrt ((m + kq) * (m - kp) * u ^ 2 +
      (m - kq) * (m + kp) * v ^ 2 +
      (m - kq) * (m - kp) * u * v * coth (u / 2) * coth (v / 2) +
      (m + kq) * (m + kp) * u * v * Real.tanh (u / 2) * Real.tanh (v / 2)) / 2

example : ySqueeze 1 = (1 : ModeMatrix Two) := by
  simpa only [ySqueeze, Real.sqrt_one, inv_one, one_smul] using
    (Matrix.fromBlocks_one :
      Matrix.fromBlocks (1 : Matrix Two Two ℝ) 0 0 (1 : Matrix Two Two ℝ) = 1)

/-- PDF p.11, (42): one-variable objective, with u the positive finite EM
normal parameter. This is not used to define the operator divergence or GREE. -/
def tmstObjective (m k u : ℝ) : ℝ :=
  -2 * Real.log (2 * Real.sinh (u / 2)) +
    u / 2 * ((m - k) * coth (u / 2) + (m + k) * Real.tanh (u / 2))

/-- PDF pp.9,11: a symmetric TMST border with equal normal eigenvalues. -/
def tmstBorderCM (u : ℝ) : ModeMatrix Two :=
  let g := coth (u / 2) / 2
  symmetricCM (2 * g ^ 2 + 1 / 2) (2 * g ^ 2 - 1 / 2) (2 * g ^ 2 - 1 / 2)

end QAlgorithms.Papers.Independent.Intakenext100.P46
Source
Gaussian relative entropy of entanglement, arXiv:quant-ph/0402109v3, https://arxiv.org/abs/quant-ph/0402109v3
Read-back

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

QAlgorithms.Papers.Independent.Intakenext100.P46.CMEM

For any finite type III, including the empty type, and real matrices α,M\alpha,Mα,M indexed by I⊔II\sqcup II⊔I, this property holds exactly when there exist a real matrix SSS on I⊔II\sqcup II⊔I and numbers gj>12g_j>\tfrac12gj​>21​ for all j∈Ij\in Ij∈I such that, writing J=(01−10)J=\begin{pmatrix}0&1\\-1&0\end{pmatrix}J=(0−1​10​) with identity blocks indexed by III, the matrix SSS is invertible, SJST=JSJS^{\mathsf T}=JSJST=J, and

α=S diag⁡(g,g) ST,M=(ST)−1diag⁡ ⁣((log⁡2gj+12gj−1)j∈I,(log⁡2gj+12gj−1)j∈I)S−1.\alpha=S\,\operatorname{diag}(g,g)\,S^{\mathsf T}, \qquad M=(S^{\mathsf T})^{-1} \operatorname{diag}\!\left( \left(\log\frac{2g_j+1}{2g_j-1}\right)_{j\in I}, \left(\log\frac{2g_j+1}{2g_j-1}\right)_{j\in I} \right)S^{-1}.α=Sdiag(g,g)ST,M=(ST)−1diag((log2gj​−12gj​+1​)j∈I​,(log2gj​−12gj​+1​)j∈I​)S−1.

Here diag⁡(g,g)\operatorname{diag}(g,g)diag(g,g) has the entries gjg_jgj​ on both diagonal blocks and zero elsewhere. If III is empty, the conditions on the gjg_jgj​ are vacuous and all matrices are empty matrices.

QAlgorithms.Papers.Independent.Intakenext100.P46.CenteredGaussianKernel

For any finite type III, including the empty type, let L=I→NL=I\to\mathbb NL=I→N, let R:L×L→CR:L\times L\to\mathbb CR:L×L→C be an arbitrary kernel, and let α\alphaα be a real matrix indexed by I⊔II\sqcup II⊔I. This property requires that RRR admit an expansion Ruv=∑rar(u)br(v)‾R_{uv}=\sum_r a_r(u)\overline{b_r(v)}Ruv​=∑r​ar​(u)br​(v)​ with ar,br∈ℓ2(L)a_r,b_r\in\ell^2(L)ar​,br​∈ℓ2(L) and ∑r∥ar∥∥br∥<∞\sum_r\|a_r\|\|b_r\|<\infty∑r​∥ar​∥∥br​∥<∞, that every finite quadratic form ∑u,v∈Fzu‾Ruvzv\sum_{u,v\in F}\overline{z_u}R_{uv}z_v∑u,v∈F​zu​​Ruv​zv​ be real and nonnegative, and that (Re⁡Ruu)u(\operatorname{Re}R_{uu})_u(ReRuu​)u​ be summable with sum 111. It also requires αT=α\alpha^{\mathsf T}=\alphaαT=α and α−i2J\alpha-\tfrac i2Jα−2i​J positive semidefinite, with J=(01−10)J=\begin{pmatrix}0&1\\-1&0\end{pmatrix}J=(0−1​10​). For every real vector ξ\xiξ on I⊔II\sqcup II⊔I, define

zj=−ξq,j+iξp,j2,D(z)rs=e−∣z∣2/2r! s!∑h=0min⁡(r,s)zr−h(−z‾)s−hh!(r−h)!(s−h)!,Wξ(u,v)=∏j∈ID(zj)uj,vj.z_j=-\frac{\xi_{\mathrm q,j}+i\xi_{\mathrm p,j}}{\sqrt2}, \quad D(z)_{rs}=e^{-|z|^2/2}\sqrt{r!\,s!} \sum_{h=0}^{\min(r,s)} \frac{z^{r-h}(-\overline z)^{s-h}}{h!(r-h)!(s-h)!}, \quad W_\xi(u,v)=\prod_{j\in I}D(z_j)_{u_j,v_j}.zj​=−2​ξq,j​+iξp,j​​,D(z)rs​=e−∣z∣2/2r!s!​h=0∑min(r,s)​h!(r−h)!(s−h)!zr−h(−z)s−h​,Wξ​(u,v)=j∈I∏​D(zj​)uj​,vj​​.

Then (RuvWξ(v,u))v(R_{uv}W_\xi(v,u))_v(Ruv​Wξ​(v,u))v​ must be summable for every uuu, its sums must be summable over uuu, and

∑u∑vRuvWξ(v,u)=exp⁡ ⁣(−12ξTJTαJξ).\sum_u\sum_v R_{uv}W_\xi(v,u) =\exp\!\left(-\frac12\xi^{\mathsf T}J^{\mathsf T}\alpha J\xi\right).u∑​v∑​Ruv​Wξ​(v,u)=exp(−21​ξTJTαJξ).

For empty III, the occupation space has one element and the product defining WξW_\xiWξ​ is 111.

QAlgorithms.Papers.Independent.Intakenext100.P46.CovBorder

For an arbitrary real 4×44\times44×4 matrix α\alphaα, indexed in the order qA,qB,pA,pBq_A,q_B,p_A,p_BqA​,qB​,pA​,pB​, this defines the existence of a state kernel σ\sigmaσ on N2\mathbb N^2N2 satisfying G(σ;0,α)G(\sigma;0,\alpha)G(σ;0,α), which is separable and for which there exist d∈R4d\in\mathbb R^4d∈R4 and a real 4×44\times44×4 matrix α′\alpha'α′ satisfying G(σ;d,α′)G(\sigma;d,\alpha')G(σ;d,α′), such that for every real ε>0\varepsilon>0ε>0 there exist a state kernel τ\tauτ on N2\mathbb N^2N2 and a real 4×44\times44×4 matrix β\betaβ with

G(τ;d,β),τ not separable,∣βij−αij′∣<εfor every i,j.G(\tau;d,\beta),\qquad \tau\ \text{not separable},\qquad |\beta_{ij}-\alpha'_{ij}|<\varepsilon\quad\text{for every }i,j.G(τ;d,β),τ not separable,∣βij​−αij′​∣<εfor every i,j.

Here a state kernel RRR on LLL has an expansion Rij=∑nun(i)vn(j)‾R_{ij}=\sum_n u_n(i)\overline{v_n(j)}Rij​=∑n​un​(i)vn​(j)​ with un,vn∈ℓ2(L;C)u_n,v_n\in\ell^2(L;\mathbb C)un​,vn​∈ℓ2(L;C) and ∑n∥un∥∥vn∥<∞\sum_n\|u_n\|\|v_n\|<\infty∑n​∥un​∥∥vn​∥<∞, has real nonnegative finite quadratic forms, and has summable real diagonal with sum 111. With J=\begin{psmallmatrix}0&I_2\\-I_2&0\end{psmallmatrix}, G(R;d,γ)G(R;d,\gamma)G(R;d,γ) means that γT=γ\gamma^{\mathsf T}=\gammaγT=γ, γ−i2J\gamma-\frac{i}{2}Jγ−2i​J is positive semidefinite, and for every ξ∈R4\xi\in\mathbb R^4ξ∈R4 the inner series and resulting outer series below are summable and satisfy

∑u∑vRuvWξ(v,u)=exp⁡ ⁣(−12ξTJTγJξ+iξTJd).\sum_u\sum_v R_{uv}W_\xi(v,u) =\exp\!\left(-\tfrac12\xi^{\mathsf T}J^{\mathsf T}\gamma J\xi +i\xi^{\mathsf T}Jd\right).u∑​v∑​Ruv​Wξ​(v,u)=exp(−21​ξTJTγJξ+iξTJd).

The entries of WξW_\xiWξ​ are Wξ(u,v)=DuAvA(zA)DuBvB(zB)W_\xi(u,v)=D_{u_Av_A}(z_A)D_{u_Bv_B}(z_B)Wξ​(u,v)=DuA​vA​​(zA​)DuB​vB​​(zB​), where zj=−(ξqj+iξpj)/2z_j=-(\xi_{q_j}+i\xi_{p_j})/\sqrt2zj​=−(ξqj​​+iξpj​​)/2​ and

Dmn(z)=e−∣z∣2/2m! n!∑h=0min⁡(m,n)zm−h(−z‾)n−hh!(m−h)!(n−h)!.D_{mn}(z)=e^{-|z|^2/2}\sqrt{m!\,n!} \sum_{h=0}^{\min(m,n)} \frac{z^{m-h}(-\overline z)^{n-h}}{h!(m-h)!(n-h)!}.Dmn​(z)=e−∣z∣2/2m!n!​h=0∑min(m,n)​h!(m−h)!(n−h)!zm−h(−z)n−h​.

A kernel is separable precisely when, for every positive δ\deltaδ, it differs by trace size less than δ\deltaδ from some finite convex combination of products Ah⊗BhA_h\otimes B_hAh​⊗Bh​ of state kernels on N\mathbb NN, with entries (Ah⊗Bh)(i,j),(i′,j′)=(Ah)ii′(Bh)jj′(A_h\otimes B_h)_{(i,j),(i',j')}=(A_h)_{ii'}(B_h)_{jj'}(Ah​⊗Bh​)(i,j),(i′,j′)​=(Ah​)ii′​(Bh​)jj′​. Trace size is the infimum in [0,∞][0,\infty][0,∞] of the nuclear sums ∑n∥un∥∥vn∥\sum_n\|u_n\|\|v_n\|∑n​∥un​∥∥vn​∥ for expansions of the difference, and is +∞+\infty+∞ if no expansion exists. The singleton occupation registers are identified with N\mathbb NN; the existential witnesses d,α′d,\alpha'd,α′ occur separately from the initial centered representation with parameter α\alphaα.

QAlgorithms.Papers.Independent.Intakenext100.P46.CovSeparable

For an arbitrary real 4×44\times44×4 matrix α\alphaα, indexed in the order qA,qB,pA,pBq_A,q_B,p_A,p_BqA​,qB​,pA​,pB​, this defines the assertion that there exists a state coefficient kernel σ\sigmaσ, indexed by N×N\mathbb N\times\mathbb NN×N, which is centered Gaussian with covariance parameter α\alphaα and is separable, with the following literal meanings. A state kernel RRR on a label set LLL admits sequences un,vn∈ℓ2(L;C)u_n,v_n\in\ell^2(L;\mathbb C)un​,vn​∈ℓ2(L;C) such that ∑n∥un∥∥vn∥<∞\sum_n\|u_n\|\|v_n\|<\infty∑n​∥un​∥∥vn​∥<∞ and Rij=∑nun(i)vn(j)‾R_{ij}=\sum_n u_n(i)\overline{v_n(j)}Rij​=∑n​un​(i)vn​(j)​; every finite quadratic form ∑i,j∈szi‾Rijzj\sum_{i,j\in s}\overline{z_i}R_{ij}z_j∑i,j∈s​zi​​Rij​zj​ is real and nonnegative; and its real diagonal is summable with sum 111. Write J=\begin{psmallmatrix}0&I_2\\-I_2&0\end{psmallmatrix}. The Gaussian condition requires αT=α\alpha^{\mathsf T}=\alphaαT=α, α−i2J\alpha-\frac{i}{2}Jα−2i​J positive semidefinite, and, for every ξ∈R4\xi\in\mathbb R^4ξ∈R4, summability of each inner series and of the resulting outer series in

∑u∑vσuvWξ(v,u)=exp⁡ ⁣(−12ξTJTαJξ).\sum_u\sum_v \sigma_{uv}W_\xi(v,u) =\exp\!\left(-\tfrac12\xi^{\mathsf T}J^{\mathsf T}\alpha J\xi\right).u∑​v∑​σuv​Wξ​(v,u)=exp(−21​ξTJTαJξ).

Here u,v∈N2u,v\in\mathbb N^2u,v∈N2, Wξ(u,v)=DuAvA(zA)DuBvB(zB)W_\xi(u,v)=D_{u_Av_A}(z_A)D_{u_Bv_B}(z_B)Wξ​(u,v)=DuA​vA​​(zA​)DuB​vB​​(zB​), zj=−(ξqj+iξpj)/2z_j=-(\xi_{q_j}+i\xi_{p_j})/\sqrt2zj​=−(ξqj​​+iξpj​​)/2​, and

Dmn(z)=e−∣z∣2/2m! n!∑h=0min⁡(m,n)zm−h(−z‾)n−hh!(m−h)!(n−h)!.D_{mn}(z)=e^{-|z|^2/2}\sqrt{m!\,n!} \sum_{h=0}^{\min(m,n)} \frac{z^{m-h}(-\overline z)^{n-h}}{h!(m-h)!(n-h)!}.Dmn​(z)=e−∣z∣2/2m!n!​h=0∑min(m,n)​h!(m−h)!(n−h)!zm−h(−z)n−h​.

Separability means that, for every ε>0\varepsilon>0ε>0, there is a finite convex combination ∑h=1qwhAh⊗Bh\sum_{h=1}^q w_h A_h\otimes B_h∑h=1q​wh​Ah​⊗Bh​, where wh≥0w_h\geq0wh​≥0, ∑hwh=1\sum_h w_h=1∑h​wh​=1, and Ah,BhA_h,B_hAh​,Bh​ are state kernels on N\mathbb NN, whose difference from σ\sigmaσ has trace size less than ε\varepsilonε. The tensor product has entries (A⊗B)(i,j),(i′,j′)=Aii′Bjj′(A\otimes B)_{(i,j),(i',j')}=A_{ii'}B_{jj'}(A⊗B)(i,j),(i′,j′)​=Aii′​Bjj′​; trace size is the infimum, in [0,∞][0,\infty][0,∞], of ∑n∥un∥∥vn∥\sum_n\|u_n\|\|v_n\|∑n​∥un​∥∥vn​∥ over all nuclear expansions of the difference of the form just specified, with infimum +∞+\infty+∞ if there is none. The two occupation labels here identify functions from a singleton to N\mathbb NN with their unique values; no separate physicality hypothesis on α\alphaα is imposed outside the existential condition.

QAlgorithms.Papers.Independent.Intakenext100.P46.GaussianBorder

Let A,BA,BA,B be arbitrary finite types, possibly empty, and L=(A→N)×(B→N)L=(A\to\mathbb N)\times(B\to\mathbb N)L=(A→N)×(B→N). States on a label set KKK mean kernels RRR with expansions Ruv=∑rar(u)br(v)‾R_{uv}=\sum_r a_r(u)\overline{b_r(v)}Ruv​=∑r​ar​(u)br​(v)​, where ar,br∈ℓ2(K)a_r,b_r\in\ell^2(K)ar​,br​∈ℓ2(K) and ∑r∥ar∥∥br∥<∞\sum_r\|a_r\|\|b_r\|<\infty∑r​∥ar​∥∥br​∥<∞, with real nonnegative finite quadratic forms and a summable real diagonal of sum 111; write ν(X)\nu(X)ν(X) for the extended nonnegative infimum of these expansion sums for XXX, taking ν(X)=∞\nu(X)=\inftyν(X)=∞ when no expansion exists. For a state σ\sigmaσ on LLL, the property first requires that, for every ε>0\varepsilon>0ε>0, its coefficients have ν\nuν-distance less than ε\varepsilonε from some finite convex combination of product kernels of states on the two occupation spaces. It then requires the existence of one real quadrature vector ddd and one real matrix α\alphaα such that σ\sigmaσ satisfies the Gaussian conditions below, and for every real ε>0\varepsilon>0ε>0 there exist another state τ\tauτ on LLL and a real matrix β\betaβ satisfying those same Gaussian conditions with the same ddd, such that τ\tauτ does not have the stated approximation property and ∣βij−αij∣<ε|\beta_{ij}-\alpha_{ij}|<\varepsilon∣βij​−αij​∣<ε for every pair of quadrature indices i,ji,ji,j. The Gaussian conditions for a kernel RRR, vector ddd, and matrix aaa are aT=aa^{\mathsf T}=aaT=a, a−i2Ja-\tfrac i2Ja−2i​J positive semidefinite, and, for every real quadrature vector ξ\xiξ,

∑u∑vRuvWξ(v,u)=exp⁡ ⁣(−12ξTJTaJξ+iξTJd),J=(01−10),\sum_u\sum_v R_{uv}W_\xi(v,u) =\exp\!\left(-\frac12\xi^{\mathsf T}J^{\mathsf T}aJ\xi+i\xi^{\mathsf T}Jd\right), \qquad J=\begin{pmatrix}0&1\\-1&0\end{pmatrix},u∑​v∑​Ruv​Wξ​(v,u)=exp(−21​ξTJTaJξ+iξTJd),J=(0−1​10​),

with summability over vvv for each uuu and summability of those sums over uuu. Here Wξ(u,v)W_\xi(u,v)Wξ​(u,v) is the product over the AAA and BBB modes of D(zi)ui,viD(z_i)_{u_i,v_i}D(zi​)ui​,vi​​, where

zi=−ξq,i+iξp,i2,D(z)rs=e−∣z∣2/2r! s!∑h=0min⁡(r,s)zr−h(−z‾)s−hh!(r−h)!(s−h)!.z_i=-\frac{\xi_{\mathrm q,i}+i\xi_{\mathrm p,i}}{\sqrt2}, \qquad D(z)_{rs}=e^{-|z|^2/2}\sqrt{r!\,s!} \sum_{h=0}^{\min(r,s)} \frac{z^{r-h}(-\overline z)^{s-h}}{h!(r-h)!(s-h)!}.zi​=−2​ξq,i​+iξp,i​​,D(z)rs​=e−∣z∣2/2r!s!​h=0∑min(r,s)​h!(r−h)!(s−h)!zr−h(−z)s−h​.

All products over empty mode sets equal 111; the required nearby nonseparable states remain part of the property even in such cases.

QAlgorithms.Papers.Independent.Intakenext100.P46.GaussianEM

For any finite type III, including the empty type, and real matrices α,M\alpha,Mα,M on I⊔II\sqcup II⊔I, put J=(01−10)J=\begin{pmatrix}0&1\\-1&0\end{pmatrix}J=(0−1​10​) and L=I→NL=I\to\mathbb NL=I→N. This property requires, first, the existence of an invertible real matrix S0S_0S0​ and numbers gj>12g_j>\tfrac12gj​>21​ such that

S0JS0T=J,α=S0diag⁡(g,g)S0T,M=(S0T)−1diag⁡(h,h)S0−1,hj=log⁡2gj+12gj−1.S_0JS_0^{\mathsf T}=J,\qquad \alpha=S_0\operatorname{diag}(g,g)S_0^{\mathsf T},\qquad M=(S_0^{\mathsf T})^{-1}\operatorname{diag}(h,h)S_0^{-1}, \quad h_j=\log\frac{2g_j+1}{2g_j-1}.S0​JS0T​=J,α=S0​diag(g,g)S0T​,M=(S0T​)−1diag(h,h)S0−1​,hj​=log2gj​−12gj​+1​.

Second, it requires the following implication for every complex kernel RRR on LLL. Suppose Ruv=∑rar(u)br(v)‾R_{uv}=\sum_r a_r(u)\overline{b_r(v)}Ruv​=∑r​ar​(u)br​(v)​ for sequences in ℓ2(L)\ell^2(L)ℓ2(L) with ∑r∥ar∥∥br∥<∞\sum_r\|a_r\|\|b_r\|<\infty∑r​∥ar​∥∥br​∥<∞, every finite quadratic form of RRR is real and nonnegative, its real diagonal is summable with sum 111, αT=α\alpha^{\mathsf T}=\alphaαT=α, and α−i2J\alpha-\tfrac i2Jα−2i​J is positive semidefinite. Suppose also that for every real quadrature vector ξ\xiξ, each family (RuvWξ(v,u))v(R_{uv}W_\xi(v,u))_v(Ruv​Wξ​(v,u))v​ and the family of its sums over uuu are summable, with

∑u∑vRuvWξ(v,u)=exp⁡ ⁣(−12ξTJTαJξ).\sum_u\sum_v R_{uv}W_\xi(v,u) =\exp\!\left(-\frac12\xi^{\mathsf T}J^{\mathsf T}\alpha J\xi\right).u∑​v∑​Ruv​Wξ​(v,u)=exp(−21​ξTJTαJξ).

Then there must exist an invertible real matrix SSS, numbers mj>0m_j>0mj​>0, and a surjective complex-linear isometry UUU of ℓ2(L)\ell^2(L)ℓ2(L) such that

SJST=J,CU−1(Wξ)=W−JSTJξ for every ξ,R=CU(Tm),M=(ST)−1diag⁡(m,m)S−1.SJS^{\mathsf T}=J,\qquad \mathcal C_{U^{-1}}(W_\xi)=W_{-JS^{\mathsf T}J\xi}\ \text{for every }\xi,\qquad R=\mathcal C_U(T_m),\qquad M=(S^{\mathsf T})^{-1}\operatorname{diag}(m,m)S^{-1}.SJST=J,CU−1​(Wξ​)=W−JSTJξ​ for every ξ,R=CU​(Tm​),M=(ST)−1diag(m,m)S−1.

Here Tm(u,v)=0T_m(u,v)=0Tm​(u,v)=0 for u≠vu\ne vu=v and Tm(u,u)=∏j(1−e−mj)e−mjujT_m(u,u)=\prod_j(1-e^{-m_j})e^{-m_j u_j}Tm​(u,u)=∏j​(1−e−mj​)e−mj​uj​; writing KV(u,a)=(Vea)(u)K_V(u,a)=(Ve_a)(u)KV​(u,a)=(Vea​)(u), the conjugation is the nested total sum CV(X)uv=∑aKV(u,a)(∑bXabKV(v,b)‾)\mathcal C_V(X)_{uv}=\sum_aK_V(u,a)\bigl(\sum_bX_{ab}\overline{K_V(v,b)}\bigr)CV​(X)uv​=∑a​KV​(u,a)(∑b​Xab​KV​(v,b)​). The kernels in these conditions are Wξ(u,v)=∏jD(zj)uj,vjW_\xi(u,v)=\prod_jD(z_j)_{u_j,v_j}Wξ​(u,v)=∏j​D(zj​)uj​,vj​​, where

zj=−ξq,j+iξp,j2,D(z)rs=e−∣z∣2/2r! s!∑h=0min⁡(r,s)zr−h(−z‾)s−hh!(r−h)!(s−h)!.z_j=-\frac{\xi_{\mathrm q,j}+i\xi_{\mathrm p,j}}{\sqrt2}, \qquad D(z)_{rs}=e^{-|z|^2/2}\sqrt{r!\,s!} \sum_{h=0}^{\min(r,s)} \frac{z^{r-h}(-\overline z)^{s-h}}{h!(r-h)!(s-h)!}.zj​=−2​ξq,j​+iξp,j​​,D(z)rs​=e−∣z∣2/2r!s!​h=0∑min(r,s)​h!(r−h)!(s−h)!zr−h(−z)s−h​.

The witnesses S,m,US,m,US,m,U may depend on RRR and need not coincide with the first witnesses S0,gS_0,gS0​,g; the universal implication does not itself assert that any kernel satisfying its antecedent exists. Empty products are 111.

QAlgorithms.Papers.Independent.Intakenext100.P46.GroupedPhase

For any type III, this abbreviates the disjoint union I⊔II\sqcup II⊔I: each element consists of an element of III together with a tag identifying the first or second copy. The two copies remain distinct even when their underlying elements agree. If III is empty, this type is empty.

QAlgorithms.Papers.Independent.Intakenext100.P46.HasFiniteEM

For any finite type III, possibly empty, put L=I→NL=I\to\mathbb NL=I→N, let RRR be a complex kernel on LLL, and let MMM be a real matrix on I⊔II\sqcup II⊔I. This property first requires that RRR have an expansion Ruv=∑rar(u)br(v)‾R_{uv}=\sum_r a_r(u)\overline{b_r(v)}Ruv​=∑r​ar​(u)br​(v)​ with ar,br∈ℓ2(L)a_r,b_r\in\ell^2(L)ar​,br​∈ℓ2(L) and ∑r∥ar∥∥br∥<∞\sum_r\|a_r\|\|b_r\|<\infty∑r​∥ar​∥∥br​∥<∞, that all its finite quadratic forms be real and nonnegative, and that its real diagonal be summable with sum 111. It then requires a real matrix SSS on I⊔II\sqcup II⊔I, numbers mj>0m_j>0mj​>0 for every j∈Ij\in Ij∈I, and a surjective complex-linear isometry U:ℓ2(L)→ℓ2(L)U:\ell^2(L)\to\ell^2(L)U:ℓ2(L)→ℓ2(L). They must satisfy invertibility of SSS, SJST=JSJS^{\mathsf T}=JSJST=J for J=(01−10)J=\begin{pmatrix}0&1\\-1&0\end{pmatrix}J=(0−1​10​), and

CU−1(Wξ)=W−JSTJξfor every real ξ,R=CU(Tm),M=(ST)−1diag⁡(m,m)S−1.\mathcal C_{U^{-1}}(W_\xi)=W_{-JS^{\mathsf T}J\xi}\quad\text{for every real }\xi, \qquad R=\mathcal C_U(T_m), \qquad M=(S^{\mathsf T})^{-1}\operatorname{diag}(m,m)S^{-1}.CU−1​(Wξ​)=W−JSTJξ​for every real ξ,R=CU​(Tm​),M=(ST)−1diag(m,m)S−1.

Here Tm(u,v)T_m(u,v)Tm​(u,v) is zero off the diagonal and equals ∏j(1−e−mj)e−mjuj\prod_j(1-e^{-m_j})e^{-m_j u_j}∏j​(1−e−mj​)e−mj​uj​ on the diagonal. To specify the conjugations literally, let KV(u,a)=(Vea)(u)K_V(u,a)=(Ve_a)(u)KV​(u,a)=(Vea​)(u) for the coordinate vector ea∈ℓ2(L)e_a\in\ell^2(L)ea​∈ℓ2(L), and set

CV(X)uv=∑aKV(u,a)(∑bXabKV(v,b)‾).\mathcal C_V(X)_{uv} =\sum_a K_V(u,a)\left(\sum_b X_{ab}\overline{K_V(v,b)}\right).CV​(X)uv​=a∑​KV​(u,a)(b∑​Xab​KV​(v,b)​).

Also Wξ(u,v)=∏jD(zj)uj,vjW_\xi(u,v)=\prod_jD(z_j)_{u_j,v_j}Wξ​(u,v)=∏j​D(zj​)uj​,vj​​, where zj=−(ξq,j+iξp,j)/2z_j=-(\xi_{\mathrm q,j}+i\xi_{\mathrm p,j})/\sqrt2zj​=−(ξq,j​+iξp,j​)/2​ and

D(z)rs=e−∣z∣2/2r! s!∑h=0min⁡(r,s)zr−h(−z‾)s−hh!(r−h)!(s−h)!.D(z)_{rs}=e^{-|z|^2/2}\sqrt{r!\,s!} \sum_{h=0}^{\min(r,s)} \frac{z^{r-h}(-\overline z)^{s-h}}{h!(r-h)!(s-h)!}.D(z)rs​=e−∣z∣2/2r!s!​h=0∑min(r,s)​h!(r−h)!(s−h)!zr−h(−z)s−h​.

The nested sums defining CV\mathcal C_VCV​ are the total infinite-sum operations; no separate summability clauses for them occur in this property. Empty products are 111, and conditions indexed by empty III are vacuous.

QAlgorithms.Papers.Independent.Intakenext100.P46.HasGaussian

Let A,BA,BA,B be arbitrary finite types, possibly empty, put L=(A→N)×(B→N)L=(A\to\mathbb N)\times(B\to\mathbb N)L=(A→N)×(B→N), and let ρ\rhoρ be a state with coefficient kernel R:L×L→CR:L\times L\to\mathbb CR:L×L→C. Being a state means that there exist ar,br∈ℓ2(L)a_r,b_r\in\ell^2(L)ar​,br​∈ℓ2(L), indexed by r∈Nr\in\mathbb Nr∈N, with ∑r∥ar∥∥br∥<∞\sum_r\|a_r\|\|b_r\|<\infty∑r​∥ar​∥∥br​∥<∞ and Ruv=∑rar(u)br(v)‾R_{uv}=\sum_r a_r(u)\overline{b_r(v)}Ruv​=∑r​ar​(u)br​(v)​; every finite quadratic form ∑u,v∈Fzu‾Ruvzv\sum_{u,v\in F}\overline{z_u}R_{uv}z_v∑u,v∈F​zu​​Ruv​zv​ is real and nonnegative; and the real diagonal entries form a summable family with sum 111. For a real vector ddd on (A⊔B)⊔(A⊔B)(A\sqcup B)\sqcup(A\sqcup B)(A⊔B)⊔(A⊔B) and a real matrix α\alphaα on that same index set, the property requires αT=α\alpha^{\mathsf T}=\alphaαT=α and α−i2J\alpha-\tfrac i2Jα−2i​J positive semidefinite, where J=(01−10)J=\begin{pmatrix}0&1\\-1&0\end{pmatrix}J=(0−1​10​) on the two copies of A⊔BA\sqcup BA⊔B. To describe its remaining requirements, for z∈Cz\in\mathbb Cz∈C and r,s∈Nr,s\in\mathbb Nr,s∈N define

D(z)rs=e−∣z∣2/2r! s!∑h=0min⁡(r,s)zr−h(−z‾)s−hh!(r−h)!(s−h)!.D(z)_{rs} =e^{-|z|^2/2}\sqrt{r!\,s!} \sum_{h=0}^{\min(r,s)} \frac{z^{r-h}(-\overline z)^{s-h}}{h!(r-h)!(s-h)!}.D(z)rs​=e−∣z∣2/2r!s!​h=0∑min(r,s)​h!(r−h)!(s−h)!zr−h(−z)s−h​.

For every real quadrature vector ξ\xiξ, set zi=−(ξq,i+iξp,i)/2z_i=-(\xi_{\mathrm q,i}+i\xi_{\mathrm p,i})/\sqrt2zi​=−(ξq,i​+iξp,i​)/2​ and

Wξ(u,v)=∏a∈AD(za)uA(a),vA(a)∏b∈BD(zb)uB(b),vB(b).W_\xi(u,v)= \prod_{a\in A}D(z_a)_{u_A(a),v_A(a)} \prod_{b\in B}D(z_b)_{u_B(b),v_B(b)}.Wξ​(u,v)=a∈A∏​D(za​)uA​(a),vA​(a)​b∈B∏​D(zb​)uB​(b),vB​(b)​.

Then, for every ξ\xiξ and every u∈Lu\in Lu∈L, the family (RuvWξ(v,u))v∈L(R_{uv}W_\xi(v,u))_{v\in L}(Ruv​Wξ​(v,u))v∈L​ must be summable, the family of its sums indexed by uuu must also be summable, and

∑u∑vRuvWξ(v,u)=exp⁡ ⁣(−12ξTJTαJξ+i ξTJd).\sum_u\sum_v R_{uv}W_\xi(v,u) =\exp\!\left(-\frac12\xi^{\mathsf T}J^{\mathsf T}\alpha J\xi +i\,\xi^{\mathsf T}Jd\right).u∑​v∑​Ruv​Wξ​(v,u)=exp(−21​ξTJTαJξ+iξTJd).

Products over empty mode sets equal 111; no nonempty-mode hypothesis is imposed.

QAlgorithms.Papers.Independent.Intakenext100.P46.IsGaussianDensity

For arbitrary finite types A,BA,BA,B, including empty ones, let L=(A→N)×(B→N)L=(A\to\mathbb N)\times(B\to\mathbb N)L=(A→N)×(B→N) and let ρ\rhoρ be a state with coefficients RuvR_{uv}Ruv​. Explicitly, RRR admits an expansion Ruv=∑rar(u)br(v)‾R_{uv}=\sum_r a_r(u)\overline{b_r(v)}Ruv​=∑r​ar​(u)br​(v)​ with ar,br∈ℓ2(L)a_r,b_r\in\ell^2(L)ar​,br​∈ℓ2(L) and ∑r∥ar∥∥br∥<∞\sum_r\|a_r\|\|b_r\|<\infty∑r​∥ar​∥∥br​∥<∞, all its finite quadratic forms are real and nonnegative, and its real diagonal is summable with sum 111. This property says that there exist a real vector ddd and a real matrix α\alphaα, both indexed by the two quadrature copies of A⊔BA\sqcup BA⊔B, such that αT=α\alpha^{\mathsf T}=\alphaαT=α, α−i2J\alpha-\tfrac i2Jα−2i​J is positive semidefinite for J=(01−10)J=\begin{pmatrix}0&1\\-1&0\end{pmatrix}J=(0−1​10​), and the following holds for every real quadrature vector ξ\xiξ. Define

D(z)rs=e−∣z∣2/2r! s!∑h=0min⁡(r,s)zr−h(−z‾)s−hh!(r−h)!(s−h)!,zi=−ξq,i+iξp,i2,D(z)_{rs} =e^{-|z|^2/2}\sqrt{r!\,s!} \sum_{h=0}^{\min(r,s)} \frac{z^{r-h}(-\overline z)^{s-h}}{h!(r-h)!(s-h)!}, \quad z_i=-\frac{\xi_{\mathrm q,i}+i\xi_{\mathrm p,i}}{\sqrt2},D(z)rs​=e−∣z∣2/2r!s!​h=0∑min(r,s)​h!(r−h)!(s−h)!zr−h(−z)s−h​,zi​=−2​ξq,i​+iξp,i​​,

and Wξ(u,v)=∏a∈AD(za)uA(a),vA(a)∏b∈BD(zb)uB(b),vB(b)W_\xi(u,v)=\prod_{a\in A}D(z_a)_{u_A(a),v_A(a)}\prod_{b\in B}D(z_b)_{u_B(b),v_B(b)}Wξ​(u,v)=∏a∈A​D(za​)uA​(a),vA​(a)​∏b∈B​D(zb​)uB​(b),vB​(b)​. Each inner family (RuvWξ(v,u))v(R_{uv}W_\xi(v,u))_v(Ruv​Wξ​(v,u))v​ is summable, the family of inner sums is summable over uuu, and

∑u∑vRuvWξ(v,u)=exp⁡ ⁣(−12ξTJTαJξ+i ξTJd).\sum_u\sum_v R_{uv}W_\xi(v,u) =\exp\!\left(-\frac12\xi^{\mathsf T}J^{\mathsf T}\alpha J\xi +i\,\xi^{\mathsf T}Jd\right).u∑​v∑​Ruv​Wξ​(v,u)=exp(−21​ξTJTαJξ+iξTJd).

The witnesses d,αd,\alphad,α are existential and are not asserted to be unique; empty products are 111.

QAlgorithms.Papers.Independent.Intakenext100.P46.IsPhysicalCovariance

For every finite type III and every real matrix α\alphaα indexed by I⊔II\sqcup II⊔I, this predicate means exactly that αT=α\alpha^{\mathsf T}=\alphaαT=α and

αC−i2(01I−1I0)\alpha_{\mathbb C}-\frac{\mathrm i}{2} \begin{pmatrix}0&\mathbf 1_I\\-\mathbf 1_I&0\end{pmatrix}αC​−2i​(0−1I​​1I​0​)

is positive semidefinite, where αC\alpha_{\mathbb C}αC​ is obtained by viewing each real entry as complex. Positive semidefiniteness requires a Hermitian matrix with nonnegative quadratic form on every complex vector. Empty III is included; both conditions then hold for its unique empty matrix.

QAlgorithms.Papers.Independent.Intakenext100.P46.IsSymplecticMatrix

For every finite type III and real matrix SSS indexed by I⊔II\sqcup II⊔I, this predicate requires both that SSS be a unit in the matrix ring—there exists a matrix TTT with ST=TS=1ST=TS=\mathbf 1ST=TS=1—and that

SΩIST=ΩI,ΩI=(01I−1I0).S\Omega_I S^{\mathsf T}=\Omega_I, \qquad \Omega_I=\begin{pmatrix}0&\mathbf 1_I\\-\mathbf 1_I&0\end{pmatrix}.SΩI​ST=ΩI​,ΩI​=(0−1I​​1I​0​).

The definition includes empty III; its unique empty matrix satisfies both requirements.

QAlgorithms.Papers.Independent.Intakenext100.P46.ModeMatrix

For any type III, this abbreviates the type of real matrices whose rows and columns are indexed by I⊔II\sqcup II⊔I, the disjoint union of two tagged copies of III. Equivalently, such a matrix is an arbitrary function (I⊔I)×(I⊔I)→R(I\sqcup I)\times(I\sqcup I)\to\mathbb R(I⊔I)×(I⊔I)→R. No finiteness, symmetry, or positivity condition is included; for empty III, there is a unique empty matrix.

QAlgorithms.Papers.Independent.Intakenext100.P46.PositiveRoots

For real numbers K,xK,xK,x, this predicate holds exactly when K≥2K\ge2K≥2 and

x=K+K2−42orx=K−K2−42.x=\sqrt{\frac{K+\sqrt{K^2-4}}2} \quad\text{or}\quad x=\sqrt{\frac{K-\sqrt{K^2-4}}2}.x=2K+K2−4​​​orx=2K−K2−4​​​.

The square roots are nonnegative real square roots. The disjunction permits either value and does not require them to be distinct; when K=2K=2K=2, both displayed values equal 111. For K<2K<2K<2, the predicate is false.

QAlgorithms.Papers.Independent.Intakenext100.P46.StateBorder

For arbitrary types I,JI,JI,J and a state σ\sigmaσ on I×JI\times JI×J, this predicate says that σ\sigmaσ lies in the topological boundary of the set of separable states: every open neighborhood of σ\sigmaσ meets both that set and its complement within the state space. A state on a label type LLL is a matrix RRR admitting a countable expansion Rij=∑rur(i)vr(j)‾R_{ij}=\sum_r u_r(i)\overline{v_r(j)}Rij​=∑r​ur​(i)vr​(j)​ with ur,vr∈ℓ2(L;C)u_r,v_r\in\ell^2(L;\mathbb C)ur​,vr​∈ℓ2(L;C) and ∑r∥ur∥∥vr∥<∞\sum_r\|u_r\|\|v_r\|<\infty∑r​∥ur​∥∥vr​∥<∞, having real nonnegative finite quadratic forms and a summable real diagonal of sum one. Define N(X)N(X)N(X) as the extended nonnegative infimum of ∑r∥ur∥∥vr∥\sum_r\|u_r\|\|v_r\|∑r​∥ur​∥∥vr​∥ over all such expansions of XXX, with value +∞+\infty+∞ when there are none. The state-space topology is generated by all sets {τ:N(τ−ρ)<ε}\{\tau:N(\tau-\rho)<\varepsilon\}{τ:N(τ−ρ)<ε}, where ρ\rhoρ is a state and ε>0\varepsilon>0ε>0. A state is separable when, for every ε>0\varepsilon>0ε>0, its coefficient matrix is within NNN-distance strictly less than ε\varepsilonε of a finite convex combination of product state matrices A⊗BA\otimes BA⊗B, whose entries are Aii′Bjj′A_{ii'}B_{jj'}Aii′​Bjj′​. The boundary predicate itself does not additionally require σ\sigmaσ to be separable. No countability or finiteness assumptions on I,JI,JI,J are imposed.

QAlgorithms.Papers.Independent.Intakenext100.P46.StrictPhysical

For every finite type III and every real matrix α\alphaα indexed by I⊔II\sqcup II⊔I, this predicate means exactly that αT=α\alpha^{\mathsf T}=\alphaαT=α and

αC−i2(01I−1I0)\alpha_{\mathbb C}-\frac{\mathrm i}{2} \begin{pmatrix}0&\mathbf 1_I\\-\mathbf 1_I&0\end{pmatrix}αC​−2i​(0−1I​​1I​0​)

is positive definite, where real entries are embedded into C\mathbb CC. Positive definiteness requires a Hermitian matrix whose quadratic form is strictly positive on every nonzero complex vector. Empty III is allowed; there are then no nonzero vectors, so the strict-positivity requirement is vacuous and the predicate holds for the empty matrix.

QAlgorithms.Papers.Independent.Intakenext100.P46.Two

This abbreviation denotes the disjoint union of two copies of the singleton set {0}\{0\}{0}, hence a two-element type whose distinct elements may be written LLL and RRR, corresponding respectively to 000 in the left and right copies.

QAlgorithms.Papers.Independent.Intakenext100.P46.TypeIIDomain

For arbitrary real numbers gA,gB,xg_A,g_B,xgA​,gB​,x, this defines the proposition

x>0and[max⁡ ⁣(gAgB,gBgA)<x  or  x<min⁡ ⁣(gAgB,gBgA)].x>0 \quad\text{and}\quad \left[ \max\!\left(\sqrt{\frac{g_A}{g_B}},\sqrt{\frac{g_B}{g_A}}\right)<x \ \ \text{or}\ \ x<\min\!\left(\sqrt{\frac{g_A}{g_B}},\sqrt{\frac{g_B}{g_A}}\right) \right].x>0and[max(gB​gA​​​,gA​gB​​​)<x  or  x<min(gB​gA​​​,gA​gB​​​)].

Both comparisons are strict. No positivity or nonzero assumption is imposed on gAg_AgA​ or gBg_BgB​; division by zero equals zero and the real square root of a nonpositive number equals zero.

QAlgorithms.Papers.Independent.Intakenext100.P46.borderCMFrom

For any real 2×22\times22×2 matrix QQQ, indexed by the disjoint union of two singleton sets, and any real numbers gA,gBg_A,g_BgA​,gB​, this defines the real 4×44\times44×4 matrix

borderCMFrom⁡(Q,gA,gB)=L(Q) diag⁡(gA,gB,gA,gB) L(Q)T,L(Q)=(Q00(QT)−1).\operatorname{borderCMFrom}(Q,g_A,g_B) =L(Q)\,\operatorname{diag}(g_A,g_B,g_A,g_B)\,L(Q)^{\mathsf T}, \qquad L(Q)=\begin{pmatrix}Q&0\\0&(Q^{\mathsf T})^{-1}\end{pmatrix}.borderCMFrom(Q,gA​,gB​)=L(Q)diag(gA​,gB​,gA​,gB​)L(Q)T,L(Q)=(Q0​0(QT)−1​).

The quadrature blocks follow the order of those two singletons. The definition assumes neither that QQQ is invertible nor that the result is a physical covariance; the matrix inverse is total and is the zero matrix when its argument is singular.

QAlgorithms.Papers.Independent.Intakenext100.P46.borderLeft

For all real numbers gA,gBg_A,g_BgA​,gB​, with no additional restrictions, this defines

borderLeft⁡(gA,gB)=(2gA2−12)(2gB2−12).\operatorname{borderLeft}(g_A,g_B) =\left(2g_A^2-\frac12\right)\left(2g_B^2-\frac12\right).borderLeft(gA​,gB​)=(2gA2​−21​)(2gB2​−21​).

QAlgorithms.Papers.Independent.Intakenext100.P46.borderRightI

For arbitrary real numbers gA,gB,r,xg_A,g_B,r,xgA​,gB​,r,x, this defines the real-valued function

borderRightI⁡(gA,gB,r,x)=sinh⁡2(2r)((x2+(x−1)2)gAgB+gA2+gB2).\operatorname{borderRightI}(g_A,g_B,r,x) =\sinh^2(2r)\left((x^2+(x^{-1})^2)g_Ag_B+g_A^2+g_B^2\right).borderRightI(gA​,gB​,r,x)=sinh2(2r)((x2+(x−1)2)gA​gB​+gA2​+gB2​).

There are no positivity or nonzero assumptions on the arguments; in particular, x=0x=0x=0 is allowed, with the convention 0−1=00^{-1}=00−1=0.

QAlgorithms.Papers.Independent.Intakenext100.P46.borderRightII

For arbitrary real numbers gA,gB,t,xg_A,g_B,t,xgA​,gB​,t,x, this defines the real-valued function

borderRightII⁡(gA,gB,t,x)=sin⁡2(2t)((x2+(x−1)2)gAgB−gA2−gB2).\operatorname{borderRightII}(g_A,g_B,t,x) =\sin^2(2t)\left((x^2+(x^{-1})^2)g_Ag_B-g_A^2-g_B^2\right).borderRightII(gA​,gB​,t,x)=sin2(2t)((x2+(x−1)2)gA​gB​−gA2​−gB2​).

There are no positivity or nonzero assumptions on the arguments; x=0x=0x=0 is included, with 0−1=00^{-1}=00−1=0.

QAlgorithms.Papers.Independent.Intakenext100.P46.bosonic

For every real number xxx, the declaration defines

bosonic⁡(x)=(x+1)log⁡(x+1)−xlog⁡x.\operatorname{bosonic}(x)=(x+1)\log(x+1)-x\log x.bosonic(x)=(x+1)log(x+1)−xlogx.

The domain is all of R\mathbb RR, with no nonnegativity assumption. The real logarithm is total, taking value 000 at zero and log⁡∣t∣\log|t|log∣t∣ at negative ttt; consequently the formula is defined at x=0x=0x=0, x=−1x=-1x=−1, and negative arguments as well.

QAlgorithms.Papers.Independent.Intakenext100.P46.congruence

For arbitrary real 4×44\times44×4 matrices SSS and α\alphaα, indexed by two position labels followed by two momentum labels, this defines

congruence⁡(S,α)=SαST.\operatorname{congruence}(S,\alpha)=S\alpha S^{\mathsf T}.congruence(S,α)=SαST.

There is no assumption that SSS is invertible or that α\alphaα is symmetric or positive.

QAlgorithms.Papers.Independent.Intakenext100.P46.coth

For every real number xxx, this defines coth⁡(x)=cosh⁡(x)/sinh⁡(x)\operatorname{coth}(x)=\cosh(x)/\sinh(x)coth(x)=cosh(x)/sinh(x). The quotient uses total real division, so its value at x=0x=0x=0 is 000.

QAlgorithms.Papers.Independent.Intakenext100.P46.covarianceObjective

For every finite type III, every pair of real matrices α,M\alpha,Mα,M indexed by I⊔II\sqcup II⊔I, and every c∈Rc\in\mathbb Rc∈R, the declaration defines

covarianceObjective⁡(α,M,c)=−log⁡c+12tr⁡(αM)=−log⁡c+12∑i,j∈I⊔IαijMji.\operatorname{covarianceObjective}(\alpha,M,c) =-\log c+\frac12\operatorname{tr}(\alpha M) =-\log c+\frac12\sum_{i,j\in I\sqcup I}\alpha_{ij}M_{ji}.covarianceObjective(α,M,c)=−logc+21​tr(αM)=−logc+21​i,j∈I⊔I∑​αij​Mji​.

No symmetry, positivity, covariance, or normalization condition is imposed on these arguments. The real logarithm is defined at every real ccc, with log⁡0=0\log0=0log0=0 and log⁡c=log⁡∣c∣\log c=\log|c|logc=log∣c∣ when c<0c<0c<0. If III is empty, the trace term is zero.

QAlgorithms.Papers.Independent.Intakenext100.P46.covarianceType

For arbitrary real numbers a,b,c1,c2a,b,c_1,c_2a,b,c1​,c2​, this defines a natural number by the ordered cases

covarianceType⁡(a,b,c1,c2)={4,a=b,1,a≠b and c1c2+c2c1<ab+ba,2,a≠b and ab+ba<c1c2+c2c1,3,a≠b and ab+ba=c1c2+c2c1.\operatorname{covarianceType}(a,b,c_1,c_2)= \begin{cases} 4,&a=b,\\ 1,&a\ne b\ \text{and }\dfrac{c_1}{c_2}+\dfrac{c_2}{c_1} <\dfrac ab+\dfrac ba,\\ 2,&a\ne b\ \text{and }\dfrac ab+\dfrac ba <\dfrac{c_1}{c_2}+\dfrac{c_2}{c_1},\\ 3,&a\ne b\ \text{and }\dfrac ab+\dfrac ba =\dfrac{c_1}{c_2}+\dfrac{c_2}{c_1}. \end{cases}covarianceType(a,b,c1​,c2​)=⎩⎨⎧​4,1,2,3,​a=b,a=b and c2​c1​​+c1​c2​​<ba​+ab​,a=b and ba​+ab​<c2​c1​​+c1​c2​​,a=b and ba​+ab​=c2​c1​​+c1​c2​​.​

All four inputs are unrestricted; every quotient uses the convention that division by zero equals zero. In particular, the first case takes precedence regardless of the quotient values.

QAlgorithms.Papers.Independent.Intakenext100.P46.densityEntropy

For any type III, put HI=ℓ2(I;C)\mathcal H_I=\ell^2(I;\mathbb C)HI​=ℓ2(I;C) and let eie_iei​ be its coordinate unit vectors. Let ρ\rhoρ be a state matrix: its entries admit a countable expansion ρij=∑rur(i)vr(j)‾\rho_{ij}=\sum_r u_r(i)\overline{v_r(j)}ρij​=∑r​ur​(i)vr​(j)​ with ur,vr∈HIu_r,v_r\in\mathcal H_Iur​,vr​∈HI​ and ∑r∥ur∥∥vr∥<∞\sum_r\|u_r\|\|v_r\|<\infty∑r​∥ur​∥∥vr​∥<∞; all its finite quadratic forms are real and nonnegative; and its real diagonal is summable with sum one. Let RRR be the chosen bounded complex-linear operator satisfying (Rej)i=ρij(Re_j)_i=\rho_{ij}(Rej​)i​=ρij​, with zero used if no such operator exists. The definition assigns

∑i∈Imax⁡ ⁣{0,Re⁡⟨ei,h(R)ei⟩}∈[0,∞],h(x)=−xlog⁡x,\sum_{i\in I} \max\!\left\{0,\operatorname{Re}\langle e_i,h(R)e_i\rangle\right\} \in[0,\infty], \qquad h(x)=-x\log x,i∈I∑​max{0,Re⟨ei​,h(R)ei​⟩}∈[0,∞],h(x)=−xlogx,

where h(R)h(R)h(R) is formed using the real continuous functional calculus for self-adjoint operators. This is an extended nonnegative sum and may be infinite. The logarithm is natural, and the scalar function has h(0)=0h(0)=0h(0)=0 under the total real-logarithm convention. No countability assumption on III is required; if III is empty, there is no state satisfying the diagonal-sum condition.

QAlgorithms.Papers.Independent.Intakenext100.P46.displacementPenalty

For every finite type III, every real matrix MMM indexed by I⊔II\sqcup II⊔I, and every real vector zzz on I⊔II\sqcup II⊔I, put

Ω=(0I−I0).\Omega=\begin{pmatrix}0&I\\-I&0\end{pmatrix}.Ω=(0−I​I0​).

The declaration defines

displacementPenalty⁡(M,z)=12∑i,j∈I⊔I(Ωz)iMij(Ωz)j.\operatorname{displacementPenalty}(M,z) =\frac12\sum_{i,j\in I\sqcup I}(\Omega z)_iM_{ij}(\Omega z)_j.displacementPenalty(M,z)=21​i,j∈I⊔I∑​(Ωz)i​Mij​(Ωz)j​.

Thus (Ωz)qj=zpj(\Omega z)_{q_j}=z_{p_j}(Ωz)qj​​=zpj​​ and (Ωz)pj=−zqj(\Omega z)_{p_j}=-z_{q_j}(Ωz)pj​​=−zqj​​. No symmetry or positivity assumption on MMM, and no condition on zzz, is imposed; the definition itself does not assert that its value is nonnegative. If III is empty, the sum is zero.

QAlgorithms.Papers.Independent.Intakenext100.P46.emFrom

For every finite type III, every real matrix SSS indexed by I⊔II\sqcup II⊔I, and every function g:I→Rg:I\to\mathbb Rg:I→R, define

ℓ(t)=log⁡ ⁣(2t+12t−1),Dg=(diag⁡(ℓ(gj))00diag⁡(ℓ(gj))).\ell(t)=\log\!\left(\frac{2t+1}{2t-1}\right), \qquad D_g=\begin{pmatrix} \operatorname{diag}(\ell(g_j))&0\\ 0&\operatorname{diag}(\ell(g_j)) \end{pmatrix}.ℓ(t)=log(2t−12t+1​),Dg​=(diag(ℓ(gj​))0​0diag(ℓ(gj​))​).

The declaration defines the real matrix

emFrom⁡(S,g)=(ST)−1DgS−1.\operatorname{emFrom}(S,g)=(S^{\mathsf T})^{-1}D_gS^{-1}.emFrom(S,g)=(ST)−1Dg​S−1.

There are no invertibility, symplecticity, or positivity assumptions on SSS or ggg. Matrix inversion and real division are total operations: a singular square matrix has inverse zero, and division by zero gives zero. The real logarithm is also total, with log⁡0=0\log 0=0log0=0 and log⁡t=log⁡∣t∣\log t=\log|t|logt=log∣t∣ for negative ttt; in particular, the formula remains defined when gj=12g_j=\tfrac12gj​=21​. The finite type III may be empty, in which case this is the unique matrix with empty index set.

QAlgorithms.Papers.Independent.Intakenext100.P46.exceptionalDelta

For arbitrary real numbers gA,gBg_A,g_BgA​,gB​, this defines

exceptionalDelta⁡(gA,gB)=(gA2−14)(gB2−14).\operatorname{exceptionalDelta}(g_A,g_B) =(g_A^2-\tfrac14)(g_B^2-\tfrac14).exceptionalDelta(gA​,gB​)=(gA2​−41​)(gB2​−41​).

No lower bounds on gA,gBg_A,g_BgA​,gB​, and no sign condition on the resulting value, are imposed.

QAlgorithms.Papers.Independent.Intakenext100.P46.exceptionalLower

For arbitrary real numbers gA,gBg_A,g_BgA​,gB​, let d=(gA2−14)(gB2−14)d=(g_A^2-\frac14)(g_B^2-\frac14)d=(gA2​−41​)(gB2​−41​). This defines a real matrix indexed by the disjoint union of two singleton sets; ordering the left singleton before the right singleton, it is

(rpow⁡(1+d/gA2,14)0rpow⁡ ⁣(d2gA2(gB2+d),14)rpow⁡ ⁣(gB2gB2+d,14)).\begin{pmatrix} \operatorname{rpow}(1+d/g_A^2,\tfrac14)&0\\[1mm] \operatorname{rpow}\!\left(\dfrac{d^2}{g_A^2(g_B^2+d)},\tfrac14\right)& \operatorname{rpow}\!\left(\dfrac{g_B^2}{g_B^2+d},\tfrac14\right) \end{pmatrix}.​rpow(1+d/gA2​,41​)rpow(gA2​(gB2​+d)d2​,41​)​0rpow(gB2​+dgB2​​,41​)​​.

Here rpow⁡\operatorname{rpow}rpow is the total real-power function used in the declaration. No positivity or invertibility hypotheses are part of the definition, and division by zero is assigned the value 000.

QAlgorithms.Papers.Independent.Intakenext100.P46.exceptionalUpper

For arbitrary real numbers gA,gBg_A,g_BgA​,gB​, let d=(gA2−14)(gB2−14)d=(g_A^2-\frac14)(g_B^2-\frac14)d=(gA2​−41​)(gB2​−41​). This defines a real matrix indexed by the disjoint union of two singleton sets; ordering the left singleton before the right singleton, it is

(rpow⁡ ⁣(gA2gA2+d,14)rpow⁡ ⁣(d2gB2(gA2+d),14)0rpow⁡(1+d/gB2,14)).\begin{pmatrix} \operatorname{rpow}\!\left(\dfrac{g_A^2}{g_A^2+d},\tfrac14\right)& \operatorname{rpow}\!\left(\dfrac{d^2}{g_B^2(g_A^2+d)},\tfrac14\right)\\[1mm] 0&\operatorname{rpow}(1+d/g_B^2,\tfrac14) \end{pmatrix}.​rpow(gA2​+dgA2​​,41​)0​rpow(gB2​(gA2​+d)d2​,41​)rpow(1+d/gB2​,41​)​​.

Here rpow⁡\operatorname{rpow}rpow is the total real-power function used in the declaration. The arguments are unrestricted, including cases with zero denominators, for which division has value 000.

QAlgorithms.Papers.Independent.Intakenext100.P46.gree

For arbitrary finite types A,BA,BA,B, possibly empty, put L=(A→N)×(B→N)L=(A\to\mathbb N)\times(B\to\mathbb N)L=(A→N)×(B→N) and let ρ\rhoρ be a state on LLL. For any label set KKK, a state here is a coefficient kernel RRR admitting an expansion Ruv=∑rar(u)br(v)‾R_{uv}=\sum_r a_r(u)\overline{b_r(v)}Ruv​=∑r​ar​(u)br​(v)​ with ar,br∈ℓ2(K)a_r,b_r\in\ell^2(K)ar​,br​∈ℓ2(K) and ∑r∥ar∥∥br∥<∞\sum_r\|a_r\|\|b_r\|<\infty∑r​∥ar​∥∥br​∥<∞, whose finite quadratic forms are real and nonnegative and whose real diagonal is summable with sum 111. Define ν(X)\nu(X)ν(X) to be the infimum, in [0,∞][0,\infty][0,∞], of ∑r∥ar∥∥br∥\sum_r\|a_r\|\|b_r\|∑r​∥ar​∥∥br​∥ over all such expansions of a kernel XXX, with value ∞\infty∞ if there is none. A state σ\sigmaσ on LLL is admitted in the infimum defining this quantity exactly when both of the following hold: its coefficients can, for every ε>0\varepsilon>0ε>0, be approximated within ν\nuν-distance strictly less than ε\varepsilonε by a finite convex combination of kernels P⊗QP\otimes QP⊗Q, where P,QP,QP,Q are states on A→NA\to\mathbb NA→N and B→NB\to\mathbb NB→N, respectively, and (P⊗Q)(uA,uB),(vA,vB)=PuAvAQuBvB(P\otimes Q)_{(u_A,u_B),(v_A,v_B)}=P_{u_Av_A}Q_{u_Bv_B}(P⊗Q)(uA​,uB​),(vA​,vB​)​=PuA​vA​​QuB​vB​​; and there exist a real quadrature vector ddd and a symmetric real matrix α\alphaα with α−i2J\alpha-\tfrac i2Jα−2i​J positive semidefinite, where J=(01−10)J=\begin{pmatrix}0&1\\-1&0\end{pmatrix}J=(0−1​10​) on the two copies of A⊔BA\sqcup BA⊔B, such that the explicitly summable characteristic series satisfies

∑u∑vσuvWξ(v,u)=exp⁡ ⁣(−12ξTJTαJξ+iξTJd)for every real ξ.\sum_u\sum_v \sigma_{uv}W_\xi(v,u) =\exp\!\left(-\frac12\xi^{\mathsf T}J^{\mathsf T}\alpha J\xi+i\xi^{\mathsf T}Jd\right) \quad\text{for every real }\xi.u∑​v∑​σuv​Wξ​(v,u)=exp(−21​ξTJTαJξ+iξTJd)for every real ξ.

Here summability is required first over vvv for each uuu, then over uuu; Wξ(u,v)W_\xi(u,v)Wξ​(u,v) is the product over all modes of D(zi)ui,viD(z_i)_{u_i,v_i}D(zi​)ui​,vi​​, with zi=−(ξq,i+iξp,i)/2z_i=-(\xi_{\mathrm q,i}+i\xi_{\mathrm p,i})/\sqrt2zi​=−(ξq,i​+iξp,i​)/2​ and

D(z)rs=e−∣z∣2/2r! s!∑h=0min⁡(r,s)zr−h(−z‾)s−hh!(r−h)!(s−h)!.D(z)_{rs}=e^{-|z|^2/2}\sqrt{r!\,s!} \sum_{h=0}^{\min(r,s)} \frac{z^{r-h}(-\overline z)^{s-h}}{h!(r-h)!(s-h)!}.D(z)rs​=e−∣z∣2/2r!s!​h=0∑min(r,s)​h!(r−h)!(s−h)!zr−h(−z)s−h​.

The value assigned is

inf⁡σ admitted sup⁡n∈Nv0,…,vn−1 orthonormal in ℓ2(L)F ⁣((⟨vi,ρ^vj⟩)ij,(⟨vi,σ^vj⟩)ij),\inf_{\sigma\ \text{admitted}}\ \sup_{\substack{n\in\mathbb N\\v_0,\ldots,v_{n-1}\text{ orthonormal in }\ell^2(L)}} F\!\left((\langle v_i,\widehat\rho v_j\rangle)_{ij}, (\langle v_i,\widehat\sigma v_j\rangle)_{ij}\right),σ admittedinf​ n∈Nv0​,…,vn−1​ orthonormal in ℓ2(L)​sup​F((⟨vi​,ρ​vj​⟩)ij​,(⟨vi​,σvj​⟩)ij​),

where R^\widehat RR is a chosen bounded operator with coefficient kernel RRR if one exists, and is 000 otherwise. For finite matrices P,QP,QP,Q, F(P,Q)=∞F(P,Q)=\inftyF(P,Q)=∞ unless ker⁡Q⊆ker⁡P\ker Q\subseteq\ker PkerQ⊆kerP; otherwise

F(P,Q)=ofReal⁡Re⁡tr⁡(f(P)−Plog⁡Q−P+Q),f(x)=xlog⁡x,F(P,Q)=\operatorname{ofReal}\operatorname{Re}\operatorname{tr} \bigl(f(P)-P\log Q-P+Q\bigr),\qquad f(x)=x\log x,F(P,Q)=ofRealRetr(f(P)−PlogQ−P+Q),f(x)=xlogx,

using continuous functional calculus, log⁡0=0\log0=0log0=0, and ofReal⁡(r)=max⁡(r,0)\operatorname{ofReal}(r)=\max(r,0)ofReal(r)=max(r,0) in [0,∞][0,\infty][0,∞]. The supremum includes n=0n=0n=0, and an empty outer infimum is ∞\infty∞; this definition does not assert attainment or finiteness.

QAlgorithms.Papers.Independent.Intakenext100.P46.groupedCCR

For every finite type III, this defines the real matrix indexed by I⊔II\sqcup II⊔I with block form

ΩI=(01I−1I0).\Omega_I=\begin{pmatrix}0&\mathbf 1_I\\-\mathbf 1_I&0\end{pmatrix}.ΩI​=(0−1I​​1I​0​).

Thus entries within either tagged copy are zero, the entry from the first copy at iii to the second copy at jjj is δij\delta_{ij}δij​, and the reverse-block entry is −δij-\delta_{ij}−δij​. Empty III is allowed and produces the empty matrix.

QAlgorithms.Papers.Independent.Intakenext100.P46.hyperbolicMixing

For every r∈Rr\in\mathbb Rr∈R, with rows and columns indexed by the two elements of Fin(1)⊔Fin(1)\mathrm{Fin}(1)\sqcup\mathrm{Fin}(1)Fin(1)⊔Fin(1) in left-then-right order, the declaration defines

hyperbolicMixing⁡(r)=(cosh⁡rsinh⁡rsinh⁡rcosh⁡r).\operatorname{hyperbolicMixing}(r) =\begin{pmatrix} \cosh r&\sinh r\\ \sinh r&\cosh r \end{pmatrix}.hyperbolicMixing(r)=(coshrsinhr​sinhrcoshr​).

There is no restriction on rrr, including r=0r=0r=0.

QAlgorithms.Papers.Independent.Intakenext100.P46.joinedCoefficients

For arbitrary types A,BA,BA,B, with no finiteness assumptions, and any complex kernel RRR indexed by pairs of occupation functions (A→N)×(B→N)(A\to\mathbb N)\times(B\to\mathbb N)(A→N)×(B→N), this defines a kernel R~\widetilde RR indexed by occupation functions on the disjoint union A⊔BA\sqcup BA⊔B. For all u,v:A⊔B→Nu,v:A\sqcup B\to\mathbb Nu,v:A⊔B→N,

R~uv=R(u∣A,u∣B),(v∣A,v∣B).\widetilde R_{uv} = R_{(u|_A,u|_B),(v|_A,v|_B)}.Ruv​=R(u∣A​,u∣B​),(v∣A​,v∣B​)​.

The restrictions use the two injections into the disjoint union. No positivity, normalization, or summability assumption on RRR is required.

QAlgorithms.Papers.Independent.Intakenext100.P46.modeEM

For every real number γ\gammaγ, this defines the real value

log⁡ ⁣(2γ+12γ−1).\log\!\left(\frac{2\gamma+1}{2\gamma-1}\right).log(2γ−12γ+1​).

There is no restriction on γ\gammaγ. Division and logarithm are total real operations: division by zero gives zero, log⁡0=0\log 0=0log0=0, and the real logarithm of a negative nonzero argument equals the logarithm of its absolute value. In particular, the defined value is zero at both γ=12\gamma=\tfrac12γ=21​ and γ=−12\gamma=-\tfrac12γ=−21​.

QAlgorithms.Papers.Independent.Intakenext100.P46.modeRotation

For every t∈Rt\in\mathbb Rt∈R, with rows and columns indexed by the two elements of Fin(1)⊔Fin(1)\mathrm{Fin}(1)\sqcup\mathrm{Fin}(1)Fin(1)⊔Fin(1) in left-then-right order, the declaration defines

modeRotation⁡(t)=(cos⁡tsin⁡t−sin⁡tcos⁡t).\operatorname{modeRotation}(t) =\begin{pmatrix} \cos t&\sin t\\ -\sin t&\cos t \end{pmatrix}.modeRotation(t)=(cost−sint​sintcost​).

The definition applies to every real ttt, with no interval or sign restriction.

QAlgorithms.Papers.Independent.Intakenext100.P46.normalDiagonal

For every finite type III and arbitrary function g:I→Rg:I\to\mathbb Rg:I→R, this defines the real matrix indexed by I⊔II\sqcup II⊔I

(diag⁡(g)00diag⁡(g)).\begin{pmatrix}\operatorname{diag}(g)&0\\0&\operatorname{diag}(g)\end{pmatrix}.(diag(g)0​0diag(g)​).

Each tagged copy has diagonal entry g(i)g(i)g(i) at iii, and all other entries are zero. No sign or lower-bound condition on ggg is imposed. Empty III gives the empty matrix.

QAlgorithms.Papers.Independent.Intakenext100.P46.normalization

For every finite type III and every function g:I→Rg:I\to\mathbb Rg:I→R, the declaration defines

normalization⁡(g)=∏j∈I2sinh⁡ ⁣(12log⁡ ⁣(2gj+12gj−1)).\operatorname{normalization}(g) =\prod_{j\in I} 2\sinh\!\left( \frac12\log\!\left(\frac{2g_j+1}{2g_j-1}\right) \right).normalization(g)=j∈I∏​2sinh(21​log(2gj​−12gj​+1​)).

No lower bound on gjg_jgj​ is assumed. Division by zero gives zero, and the total real logarithm satisfies log⁡0=0\log0=0log0=0 and log⁡t=log⁡∣t∣\log t=\log|t|logt=log∣t∣ for negative ttt; thus all real inputs are included, and a factor with gj=12g_j=\tfrac12gj​=21​ is zero. If III is empty, the product is 111.

QAlgorithms.Papers.Independent.Intakenext100.P46.operatorRelativeEntropy

For any type III, let HI=ℓ2(I;C)\mathcal H_I=\ell^2(I;\mathbb C)HI​=ℓ2(I;C), with coordinate unit vectors eie_iei​. A state on III is a matrix RRR admitting an expansion Rij=∑r=0∞ur(i)vr(j)‾R_{ij}=\sum_{r=0}^{\infty}u_r(i)\overline{v_r(j)}Rij​=∑r=0∞​ur​(i)vr​(j)​, where ur,vr∈HIu_r,v_r\in\mathcal H_Iur​,vr​∈HI​ and ∑r∥ur∥∥vr∥<∞\sum_r\|u_r\|\|v_r\|<\infty∑r​∥ur​∥∥vr​∥<∞, such that every finite quadratic form ∑i,j∈szi‾Rijzj\sum_{i,j\in s}\overline{z_i}R_{ij}z_j∑i,j∈s​zi​​Rij​zj​ is real and nonnegative, the real diagonal entries are summable, and ∑iRe⁡Rii=1\sum_i\operatorname{Re}R_{ii}=1∑i​ReRii​=1. Given two such states ρ,σ\rho,\sigmaρ,σ, choose bounded complex-linear operators R,SR,SR,S having their respective coordinate matrices; the construction assigns the zero operator if no such operator exists. The defined value is

sup⁡n∈N sup⁡v0,…,vn−1∈HI⟨vi,vj⟩=δijDfin ⁣([⟨vi,Rvj⟩]i,j,[⟨vi,Svj⟩]i,j)∈[0,∞],\sup_{n\in\mathbb N}\ \sup_{\substack{v_0,\ldots,v_{n-1}\in\mathcal H_I\\ \langle v_i,v_j\rangle=\delta_{ij}}} D_{\mathrm{fin}}\!\left( [\langle v_i,Rv_j\rangle]_{i,j}, [\langle v_i,Sv_j\rangle]_{i,j} \right)\in[0,\infty],n∈Nsup​ v0​,…,vn−1​∈HI​⟨vi​,vj​⟩=δij​​sup​Dfin​([⟨vi​,Rvj​⟩]i,j​,[⟨vi​,Svj​⟩]i,j​)∈[0,∞],

where Dfin(A,B)=+∞D_{\mathrm{fin}}(A,B)=+\inftyDfin​(A,B)=+∞ unless ker⁡B⊆ker⁡A\ker B\subseteq\ker AkerB⊆kerA; when that inclusion holds, it equals

max⁡ ⁣{0,Re⁡tr⁡(f(A)−Alog⁡B−A+B)},f(x)=xlog⁡x,\max\!\left\{0,\operatorname{Re}\operatorname{tr} \bigl(f(A)-A\log B-A+B\bigr)\right\}, \qquad f(x)=x\log x,max{0,Retr(f(A)−AlogB−A+B)},f(x)=xlogx,

with matrix functions interpreted by continuous functional calculus and log⁡\loglog the real natural logarithm. The supremum includes n=0n=0n=0, whose contribution is zero; dimensions admitting no orthonormal family contribute no values. No countability or finiteness assumption on III is imposed.

QAlgorithms.Papers.Independent.Intakenext100.P46.pairValues

For arbitrary real numbers a,ba,ba,b, this defines a real-valued function on the disjoint union of two singleton sets by assigning aaa to the left element LLL and bbb to the right element RRR.

QAlgorithms.Papers.Independent.Intakenext100.P46.ree

For arbitrary types I,JI,JI,J and a state ρ\rhoρ on I×JI\times JI×J, the definition assigns inf⁡σD(ρ,σ)∈[0,∞]\inf_\sigma D(\rho,\sigma)\in[0,\infty]infσ​D(ρ,σ)∈[0,∞], where σ\sigmaσ ranges over separable states on I×JI\times JI×J. Here a state on a label type LLL is a matrix RRR with an expansion Rij=∑rur(i)vr(j)‾R_{ij}=\sum_r u_r(i)\overline{v_r(j)}Rij​=∑r​ur​(i)vr​(j)​, where ur,vr∈ℓ2(L;C)u_r,v_r\in\ell^2(L;\mathbb C)ur​,vr​∈ℓ2(L;C) and ∑r∥ur∥∥vr∥<∞\sum_r\|u_r\|\|v_r\|<\infty∑r​∥ur​∥∥vr​∥<∞, whose finite quadratic forms are real and nonnegative and whose real diagonal is summable with sum one. To specify separability, define

N(X)=inf⁡Xij=∑rur(i)vr(j)‾∑r∥ur∥∥vr∥<∞max⁡ ⁣{0,∑r∥ur∥∥vr∥},N(X)=\inf_{\substack{X_{ij}=\sum_r u_r(i)\overline{v_r(j)}\\ \sum_r\|u_r\|\|v_r\|<\infty}} \max\!\left\{0,\sum_r\|u_r\|\|v_r\|\right\},N(X)=Xij​=∑r​ur​(i)vr​(j)​∑r​∥ur​∥∥vr​∥<∞​inf​max{0,r∑​∥ur​∥∥vr​∥},

an extended nonnegative infimum, equal to +∞+\infty+∞ if there is no such expansion. A state σ\sigmaσ is separable precisely when, for every ε>0\varepsilon>0ε>0, there is a finite convex combination YYY of product state matrices A⊗BA\otimes BA⊗B, with (A⊗B)(i,j),(i′,j′)=Aii′Bjj′(A\otimes B)_{(i,j),(i',j')}=A_{ii'}B_{jj'}(A⊗B)(i,j),(i′,j′)​=Aii′​Bjj′​, such that N(σ−Y)<εN(\sigma-Y)<\varepsilonN(σ−Y)<ε. The quantity D(ρ,σ)D(\rho,\sigma)D(ρ,σ) is the supremum over every n∈Nn\in\mathbb Nn∈N and every orthonormal family v0,…,vn−1v_0,\ldots,v_{n-1}v0​,…,vn−1​ in ℓ2(I×J;C)\ell^2(I\times J;\mathbb C)ℓ2(I×J;C) of the following value: compress the chosen bounded operators representing ρ,σ\rho,\sigmaρ,σ to matrices Aij=⟨vi,Rvj⟩A_{ij}=\langle v_i,Rv_j\rangleAij​=⟨vi​,Rvj​⟩, Bij=⟨vi,Svj⟩B_{ij}=\langle v_i,Sv_j\rangleBij​=⟨vi​,Svj​⟩; return +∞+\infty+∞ if ker⁡B⊈ker⁡A\ker B\not\subseteq\ker AkerB⊆kerA, and otherwise return max⁡{0,Re⁡tr⁡(f(A)−Alog⁡B−A+B)}\max\{0,\operatorname{Re}\operatorname{tr}(f(A)-A\log B-A+B)\}max{0,Retr(f(A)−AlogB−A+B)}, where f(x)=xlog⁡xf(x)=x\log xf(x)=xlogx and matrix functions use continuous functional calculus. A representing operator is assigned zero if none exists. The supremum includes the empty orthonormal family, and an empty set of separable comparison states gives infimum +∞+\infty+∞. No minimizing state is asserted to exist.

QAlgorithms.Papers.Independent.Intakenext100.P46.rotateBoth

For every real number ttt, this defines the real 4×44\times44×4 matrix indexed by two position labels followed by two momentum labels as

rotateBoth⁡(t)=(cos⁡tsin⁡t00−sin⁡tcos⁡t0000cos⁡tsin⁡t00−sin⁡tcos⁡t).\operatorname{rotateBoth}(t)= \begin{pmatrix} \cos t&\sin t&0&0\\ -\sin t&\cos t&0&0\\ 0&0&\cos t&\sin t\\ 0&0&-\sin t&\cos t \end{pmatrix}.rotateBoth(t)=​cost−sint00​sintcost00​00cost−sint​00sintcost​​.

QAlgorithms.Papers.Independent.Intakenext100.P46.squeezeBoth

For every real number rrr, this defines the real 4×44\times44×4 matrix indexed by two position labels followed by two momentum labels as

squeezeBoth⁡(r)=(cosh⁡rsinh⁡r00sinh⁡rcosh⁡r0000cosh⁡r−sinh⁡r00−sinh⁡rcosh⁡r).\operatorname{squeezeBoth}(r)= \begin{pmatrix} \cosh r&\sinh r&0&0\\ \sinh r&\cosh r&0&0\\ 0&0&\cosh r&-\sinh r\\ 0&0&-\sinh r&\cosh r \end{pmatrix}.squeezeBoth(r)=​coshrsinhr00​sinhrcoshr00​00coshr−sinhr​00−sinhrcoshr​​.

QAlgorithms.Papers.Independent.Intakenext100.P46.standardCM

For arbitrary real numbers a,b,c1,c2a,b,c_1,c_2a,b,c1​,c2​, this defines a real matrix indexed by two grouped copies of a two-element type. Ordering its indices as qL,qR,pL,pRq_L,q_R,p_L,p_RqL​,qR​,pL​,pR​, where L,RL,RL,R distinguish the two singleton summands, the matrix is

(ac100c1b0000a−c200−c2b).\begin{pmatrix} a&c_1&0&0\\ c_1&b&0&0\\ 0&0&a&-c_2\\ 0&0&-c_2&b \end{pmatrix}.​ac1​00​c1​b00​00a−c2​​00−c2​b​​.

No positivity or physical-covariance hypothesis is imposed on its parameters.

QAlgorithms.Papers.Independent.Intakenext100.P46.symmetricBorderQ

For arbitrary real numbers gA,gBg_A,g_BgA​,gB​, this defines a real matrix on the disjoint union of two singleton sets. In the order left singleton, right singleton, put

a=rpow⁡ ⁣(4gA2(4gB2+1)4gA2+1,14),b=rpow⁡ ⁣(4gB2+14gB2(4gA2+1),14).a=\operatorname{rpow}\!\left( \frac{4g_A^2(4g_B^2+1)}{4g_A^2+1},\tfrac14\right), \qquad b=\operatorname{rpow}\!\left( \frac{4g_B^2+1}{4g_B^2(4g_A^2+1)},\tfrac14\right).a=rpow(4gA2​+14gA2​(4gB2​+1)​,41​),b=rpow(4gB2​(4gA2​+1)4gB2​+1​,41​).

The matrix is

12(111−1)(a00b)=12(aba−b).\frac1{\sqrt2} \begin{pmatrix}1&1\\1&-1\end{pmatrix} \begin{pmatrix}a&0\\0&b\end{pmatrix} =\frac1{\sqrt2}\begin{pmatrix}a&b\\a&-b\end{pmatrix}.2​1​(11​1−1​)(a0​0b​)=2​1​(aa​b−b​).

The powers use the total real-power function; no positivity or invertibility assumption is required. In particular gB=0g_B=0gB​=0 is allowed, and the quotient defining bbb then has value 000.

QAlgorithms.Papers.Independent.Intakenext100.P46.symmetricCM

For arbitrary real numbers m,kq,kpm,k_q,k_pm,kq​,kp​, this defines the real matrix on the grouped indices qL,qR,pL,pRq_L,q_R,p_L,p_RqL​,qR​,pL​,pR​, with L,RL,RL,R the elements of a disjoint union of two singleton sets, given by

12(mkq00kqm0000m−kp00−kpm).\frac12 \begin{pmatrix} m&k_q&0&0\\ k_q&m&0&0\\ 0&0&m&-k_p\\ 0&0&-k_p&m \end{pmatrix}.21​​mkq​00​kq​m00​00m−kp​​00−kp​m​​.

There are no restrictions on the three real parameters.

QAlgorithms.Papers.Independent.Intakenext100.P46.symmetricCandidateCM

For arbitrary real numbers gA,gB,yg_A,g_B,ygA​,gB​,y, order the two singleton mode labels as A,BA,BA,B, and put

a=rpow⁡ ⁣(4gA2(4gB2+1)4gA2+1,14),b=rpow⁡ ⁣(4gB2+14gB2(4gA2+1),14),Q=12(aba−b),L=diag⁡(Q,(QT)−1),s=y−1,Ty=diag⁡(s,s,s−1,s−1).\begin{gathered} a=\operatorname{rpow}\!\left( \frac{4g_A^2(4g_B^2+1)}{4g_A^2+1},\tfrac14\right),\qquad b=\operatorname{rpow}\!\left( \frac{4g_B^2+1}{4g_B^2(4g_A^2+1)},\tfrac14\right),\\ Q=\frac1{\sqrt2}\begin{pmatrix}a&b\\a&-b\end{pmatrix}, \qquad L=\operatorname{diag}(Q,(Q^{\mathsf T})^{-1}),\\ s=\sqrt{y^{-1}},\qquad T_y=\operatorname{diag}(s,s,s^{-1},s^{-1}). \end{gathered}a=rpow(4gA2​+14gA2​(4gB2​+1)​,41​),b=rpow(4gB2​(4gA2​+1)4gB2​+1​,41​),Q=2​1​(aa​b−b​),L=diag(Q,(QT)−1),s=y−1​,Ty​=diag(s,s,s−1,s−1).​

The definition returns the real 4×44\times44×4 matrix

Ty L diag⁡(gA,gB,gA,gB) LTTyT.T_y\,L\,\operatorname{diag}(g_A,g_B,g_A,g_B)\,L^{\mathsf T}T_y^{\mathsf T}.Ty​Ldiag(gA​,gB​,gA​,gB​)LTTyT​.

There are no positivity or invertibility assumptions. Scalar division and inversion are total with 0−1=00^{-1}=00−1=0, a singular matrix has zero matrix inverse, and the real square root of a nonpositive number is 000; consequently y≤0y\le0y≤0 gives the zero matrix. For y>0y>0y>0, Ty=diag⁡(y−1/2,y−1/2,y1/2,y1/2)T_y=\operatorname{diag}(y^{-1/2},y^{-1/2},y^{1/2},y^{1/2})Ty​=diag(y−1/2,y−1/2,y1/2,y1/2).

QAlgorithms.Papers.Independent.Intakenext100.P46.symmetricGammas

For arbitrary real numbers m,kq,kpm,k_q,k_pm,kq​,kp​, this defines a real-valued function on the left and right elements L,RL,RL,R of a disjoint union of two singleton sets by

gL=(m+kq)(m−kp)2,gR=(m−kq)(m+kp)2.g_L=\frac{\sqrt{(m+k_q)(m-k_p)}}2, \qquad g_R=\frac{\sqrt{(m-k_q)(m+k_p)}}2.gL​=2(m+kq​)(m−kp​)​​,gR​=2(m−kq​)(m+kp​)​​.

No parameter restrictions are imposed; the real square-root function assigns zero to a negative argument.

QAlgorithms.Papers.Independent.Intakenext100.P46.symmetricModeBlock

For arbitrary real numbers a,c,ba,c,ba,c,b, this defines a real matrix indexed by the disjoint union of two singleton sets. In the order consisting of the left element followed by the right element, it is

(accb).\begin{pmatrix}a&c\\c&b\end{pmatrix}.(ac​cb​).

QAlgorithms.Papers.Independent.Intakenext100.P46.symmetricObjective

For arbitrary real numbers m,kq,kp,u,vm,k_q,k_p,u,vm,kq​,kp​,u,v, this defines the real number

F(m,kq,kp,u,v)=−log⁡ ⁣(2sinh⁡u2)−log⁡ ⁣(2sinh⁡v2)+12(m+kq)(m−kp)u2+(m−kq)(m+kp)v2+(m−kq)(m−kp)uvcosh⁡(u/2)sinh⁡(u/2)cosh⁡(v/2)sinh⁡(v/2)+(m+kq)(m+kp)uvtanh⁡(u/2)tanh⁡(v/2).\begin{aligned} F(m,k_q,k_p,u,v) ={}&-\log\!\left(2\sinh\frac u2\right) -\log\!\left(2\sinh\frac v2\right)\\ &+\frac12\sqrt{ (m+k_q)(m-k_p)u^2 +(m-k_q)(m+k_p)v^2 +(m-k_q)(m-k_p)uv \frac{\cosh(u/2)}{\sinh(u/2)} \frac{\cosh(v/2)}{\sinh(v/2)} +(m+k_q)(m+k_p)uv \tanh(u/2)\tanh(v/2)}. \end{aligned}F(m,kq​,kp​,u,v)=​−log(2sinh2u​)−log(2sinh2v​)+21​(m+kq​)(m−kp​)u2+(m−kq​)(m+kp​)v2+(m−kq​)(m−kp​)uvsinh(u/2)cosh(u/2)​sinh(v/2)cosh(v/2)​+(m+kq​)(m+kp​)uvtanh(u/2)tanh(v/2)​.​

There are no positivity, nonzero-denominator, or nonnegative-radicand assumptions. The real operations are total: division by zero is zero, the square root of a negative number is zero, log⁡0=0\log 0=0log0=0, and the real logarithm of a negative number is the logarithm of its absolute value.

QAlgorithms.Papers.Independent.Intakenext100.P46.symmetricSq

For arbitrary real numbers m,kq,kpm,k_q,k_pm,kq​,kp​, this defines a real matrix indexed, in order, by the left and right elements of a disjoint union of two singleton sets:

12(r+r−r+−r−),r+=rpow⁡ ⁣(m+kqm−kp,14),r−=rpow⁡ ⁣(m−kqm+kp,14).\frac1{\sqrt2} \begin{pmatrix} r_+&r_-\\ r_+&-r_- \end{pmatrix}, \qquad r_+=\operatorname{rpow}\!\left(\frac{m+k_q}{m-k_p},\frac14\right), \qquad r_-=\operatorname{rpow}\!\left(\frac{m-k_q}{m+k_p},\frac14\right).2​1​(r+​r+​​r−​−r−​​),r+​=rpow(m−kp​m+kq​​,41​),r−​=rpow(m+kp​m−kq​​,41​).

Here rpow⁡\operatorname{rpow}rpow is the total real-power function. No positivity or nonzero-denominator assumptions are imposed; the definition includes nonpositive bases and uses total real division, for which division by zero is zero.

QAlgorithms.Papers.Independent.Intakenext100.P46.symmetricValue

For any state ρ\rhoρ on pairs of occupations of two singleton mode sets and arbitrary real numbers gA,gBg_A,g_BgA​,gB​, this defines a value in [0,∞][0,\infty][0,∞] by

inf⁡y>0, σ a stateH(σ,0,αy)D(ρ,σ).\inf_{\substack{y>0,\ \sigma\text{ a state}\\ \mathcal H(\sigma,0,\alpha_y)}} \mathcal D(\rho,\sigma).y>0, σ a stateH(σ,0,αy​)​inf​D(ρ,σ).

Here a state on the occupation label set Λ≅N×N\Lambda\cong\mathbb N\times\mathbb NΛ≅N×N is a complex kernel admitting an expansion ρij=∑nun(i)vn(j)‾\rho_{ij}=\sum_nu_n(i)\overline{v_n(j)}ρij​=∑n​un​(i)vn​(j)​, with un,vn∈ℓ2(Λ)u_n,v_n\in\ell^2(\Lambda)un​,vn​∈ℓ2(Λ) and ∑n∥un∥∥vn∥<∞\sum_n\|u_n\|\|v_n\|<\infty∑n​∥un​∥∥vn​∥<∞, whose finite quadratic forms are real and nonnegative and whose real diagonal is summable with sum 111. To define the candidate covariance, put

a=rpow⁡ ⁣(4gA2(4gB2+1)4gA2+1,14),b=rpow⁡ ⁣(4gB2+14gB2(4gA2+1),14),Q=12(aba−b),L=diag⁡(Q,(QT)−1),Ty=diag⁡(y−1/2,y−1/2,y1/2,y1/2),αy=TyLdiag⁡(gA,gB,gA,gB)LTTyT,J=(0I2−I20).\begin{gathered} a=\operatorname{rpow}\!\left( \frac{4g_A^2(4g_B^2+1)}{4g_A^2+1},\tfrac14\right),\qquad b=\operatorname{rpow}\!\left( \frac{4g_B^2+1}{4g_B^2(4g_A^2+1)},\tfrac14\right),\\ Q=\frac1{\sqrt2}\begin{pmatrix}a&b\\a&-b\end{pmatrix}, \quad L=\operatorname{diag}(Q,(Q^{\mathsf T})^{-1}), \quad T_y=\operatorname{diag}(y^{-1/2},y^{-1/2},y^{1/2},y^{1/2}),\\ \alpha_y=T_y L\operatorname{diag}(g_A,g_B,g_A,g_B)L^{\mathsf T}T_y^{\mathsf T}, \qquad J=\begin{pmatrix}0&I_2\\-I_2&0\end{pmatrix}. \end{gathered}a=rpow(4gA2​+14gA2​(4gB2​+1)​,41​),b=rpow(4gB2​(4gA2​+1)4gB2​+1​,41​),Q=2​1​(aa​b−b​),L=diag(Q,(QT)−1),Ty​=diag(y−1/2,y−1/2,y1/2,y1/2),αy​=Ty​Ldiag(gA​,gB​,gA​,gB​)LTTyT​,J=(0−I2​​I2​0​).​

The condition H(σ,0,αy)\mathcal H(\sigma,0,\alpha_y)H(σ,0,αy​) requires αyT=αy\alpha_y^{\mathsf T}=\alpha_yαyT​=αy​, positive semidefiniteness of αy−i2J\alpha_y-\frac i2Jαy​−2i​J, and, for every ξ∈R4\xi\in\mathbb R^4ξ∈R4, summability of each inner series and of the outer series in

∑u∈Λ∑v∈Λσuv∏ℓ=A,BD(zℓ)vℓuℓ=exp⁡ ⁣(−12ξTJTαyJξ),zℓ=−ξqℓ+iξpℓ2,\sum_{u\in\Lambda}\sum_{v\in\Lambda} \sigma_{uv}\prod_{\ell=A,B}D(z_\ell)_{v_\ell u_\ell} =\exp\!\left(-\tfrac12\xi^{\mathsf T}J^{\mathsf T}\alpha_yJ\xi\right), \quad z_\ell=-\frac{\xi_{q_\ell}+i\xi_{p_\ell}}{\sqrt2},u∈Λ∑​v∈Λ∑​σuv​ℓ=A,B∏​D(zℓ​)vℓ​uℓ​​=exp(−21​ξTJTαy​Jξ),zℓ​=−2​ξqℓ​​+iξpℓ​​​,

where

D(z)mn=e−∣z∣2/2m! n!∑h=0min⁡(m,n)zm−h(−z‾)n−hh! (m−h)! (n−h)!.D(z)_{mn}=e^{-|z|^2/2}\sqrt{m!\,n!} \sum_{h=0}^{\min(m,n)} \frac{z^{m-h}(-\overline z)^{n-h}}{h!\,(m-h)!\,(n-h)!}.D(z)mn​=e−∣z∣2/2m!n!​h=0∑min(m,n)​h!(m−h)!(n−h)!zm−h(−z)n−h​.

For a state kernel τ\tauτ, let AτA_\tauAτ​ be a bounded operator on ℓ2(Λ)\ell^2(\Lambda)ℓ2(Λ) having that kernel if one exists, and the zero operator otherwise. The quantity D\mathcal DD is explicitly

D(ρ,σ)=sup⁡n∈N sup⁡v0,…,vn−1∈ℓ2(Λ)orthonormald ⁣([⟨vi,Aρvj⟩]ij,[⟨vi,Aσvj⟩]ij),\mathcal D(\rho,\sigma) =\sup_{n\in\mathbb N}\ \sup_{\substack{v_0,\ldots,v_{n-1}\in\ell^2(\Lambda)\\ \text{orthonormal}}} d\!\left( [\langle v_i,A_\rho v_j\rangle]_{ij}, [\langle v_i,A_\sigma v_j\rangle]_{ij}\right),D(ρ,σ)=n∈Nsup​ v0​,…,vn−1​∈ℓ2(Λ)orthonormal​sup​d([⟨vi​,Aρ​vj​⟩]ij​,[⟨vi​,Aσ​vj​⟩]ij​),

where, for the finite matrices A,BA,BA,B,

d(A,B)={max⁡ ⁣{0,Re⁡tr⁡(f(A)−Alog⁡(B)−A+B)},ker⁡B⊆ker⁡A,∞,otherwise,f(t)=tlog⁡t.d(A,B)= \begin{cases} \displaystyle \max\!\left\{0,\operatorname{Re}\operatorname{tr} \bigl(f(A)-A\log(B)-A+B\bigr)\right\}, &\ker B\subseteq\ker A,\\ \infty,&\text{otherwise}, \end{cases} \qquad f(t)=t\log t.d(A,B)={max{0,Retr(f(A)−Alog(B)−A+B)},∞,​kerB⊆kerA,otherwise,​f(t)=tlogt.

The matrix functions in this formula use continuous functional calculus. The supremum includes n=0n=0n=0, hence the empty orthonormal family. The infimum ranges over all qualifying pairs (y,σ)(y,\sigma)(y,σ), without an additional separability requirement or any restriction on gA,gBg_A,g_BgA​,gB​; it equals ∞\infty∞ if there are no qualifying pairs. Scalar quotients and matrix inverses in the candidate formula retain their total conventions, including zero scalar division and zero inverse for singular matrices.

QAlgorithms.Papers.Independent.Intakenext100.P46.symplecticLift

For every real matrix SqS_qSq​ indexed by the two elements of a disjoint union of singleton sets, this defines the real matrix on the corresponding grouped four-element index set by

(Sq00(SqT)−1).\begin{pmatrix} S_q&0\\ 0&(S_q^T)^{-1} \end{pmatrix}.(Sq​0​0(SqT​)−1​).

There is no invertibility hypothesis on SqS_qSq​; the inverse is the total matrix inverse, which is the zero matrix when SqTS_q^TSqT​ is singular.

QAlgorithms.Papers.Independent.Intakenext100.P46.thermalGibbsKernel

For any finite type III, including the empty type, and any real-valued function m:I→Rm:I\to\mathbb Rm:I→R, this defines a complex kernel on occupation functions u,v:I→Nu,v:I\to\mathbb Nu,v:I→N by

Ruv={∏j∈I(1−e−mj)e−mjuj,u=v,0,u≠v.R_{uv}= \begin{cases} \displaystyle\prod_{j\in I}(1-e^{-m_j})e^{-m_j u_j},&u=v,\\ 0,&u\ne v. \end{cases}Ruv​=⎩⎨⎧​j∈I∏​(1−e−mj​)e−mj​uj​,0,​u=v,u=v.​

The real product is regarded as a complex number. No positivity condition on mmm is imposed. If III is empty, there is one occupation function and its diagonal entry is the empty product 111; if some mj=0m_j=0mj​=0, every diagonal entry is 000.

QAlgorithms.Papers.Independent.Intakenext100.P46.tmstBorderCM

For every real uuu, put

g=cosh⁡(u/2)2sinh⁡(u/2),a=g2+14,c=g2−14.g=\frac{\cosh(u/2)}{2\sinh(u/2)},\qquad a=g^2+\frac14,\qquad c=g^2-\frac14.g=2sinh(u/2)cosh(u/2)​,a=g2+41​,c=g2−41​.

This defines the real matrix on the four coordinates (qA,qB,pA,pB)(q_A,q_B,p_A,p_B)(qA​,qB​,pA​,pB​)

(ac00ca0000a−c00−ca).\begin{pmatrix} a&c&0&0\\ c&a&0&0\\ 0&0&a&-c\\ 0&0&-c&a \end{pmatrix}.​ac00​ca00​00a−c​00−ca​​.

There is no assumption that u>0u>0u>0, that the matrix is a physical covariance, or that it has any separability property. Division by zero is zero, so at u=0u=0u=0 the definition uses g=0g=0g=0, a=14a=\tfrac14a=41​, and c=−14c=-\tfrac14c=−41​.

QAlgorithms.Papers.Independent.Intakenext100.P46.tmstObjective

For arbitrary real numbers m,k,um,k,um,k,u, this defines

T(m,k,u)=−2log⁡ ⁣(2sinh⁡u2)+u2((m−k)cosh⁡(u/2)sinh⁡(u/2)+(m+k)tanh⁡u2).T(m,k,u)= -2\log\!\left(2\sinh\frac u2\right) +\frac u2\left( (m-k)\frac{\cosh(u/2)}{\sinh(u/2)} +(m+k)\tanh\frac u2 \right).T(m,k,u)=−2log(2sinh2u​)+2u​((m−k)sinh(u/2)cosh(u/2)​+(m+k)tanh2u​).

No restrictions are imposed on the three arguments. Division by zero is zero, log⁡0=0\log0=0log0=0, and the real logarithm of a negative argument means the logarithm of its absolute value; in particular, this definition gives T(m,k,0)=0T(m,k,0)=0T(m,k,0)=0.

QAlgorithms.Papers.Independent.Intakenext100.P46.typeICM

For arbitrary real numbers gA,gB,r,xg_A,g_B,r,xgA​,gB​,r,x, define

H(s)=(cosh⁡ssinh⁡ssinh⁡scosh⁡s),B(s)=diag⁡(H(s),H(−s)),X(v)=diag⁡ ⁣(v,(v)−1,(v)−1,v).H(s)=\begin{pmatrix}\cosh s&\sinh s\\\sinh s&\cosh s\end{pmatrix}, \quad B(s)=\operatorname{diag}(H(s),H(-s)), \quad X(v)=\operatorname{diag}\!\left(\sqrt v,(\sqrt v)^{-1}, (\sqrt v)^{-1},\sqrt v\right).H(s)=(coshssinhs​sinhscoshs​),B(s)=diag(H(s),H(−s)),X(v)=diag(v​,(v​)−1,(v​)−1,v​).

This declaration defines the real 4×44\times44×4 matrix, in the order qA,qB,pA,pBq_A,q_B,p_A,p_BqA​,qB​,pA​,pB​, by

typeICM⁡(gA,gB,r,x)=[B(−r)X(x−1)] diag⁡(gA,gB,gA,gB) [B(−r)X(x−1)]T.\operatorname{typeICM}(g_A,g_B,r,x) =[B(-r)X(x^{-1})]\, \operatorname{diag}(g_A,g_B,g_A,g_B)\, [B(-r)X(x^{-1})]^{\mathsf T}.typeICM(gA​,gB​,r,x)=[B(−r)X(x−1)]diag(gA​,gB​,gA​,gB​)[B(−r)X(x−1)]T.

There are no positivity or physicality assumptions on the parameters. Reciprocals of zero are zero and real square roots of nonpositive inputs are zero; consequently the defined matrix is zero when x≤0x\leq0x≤0.

QAlgorithms.Papers.Independent.Intakenext100.P46.typeIICM

For arbitrary real numbers gA,gB,t,xg_A,g_B,t,xgA​,gB​,t,x, define

R(s)=(cos⁡ssin⁡s−sin⁡scos⁡s),T(s)=diag⁡(R(s),R(s)),X(v)=diag⁡ ⁣(v,(v)−1,(v)−1,v).R(s)=\begin{pmatrix}\cos s&\sin s\\-\sin s&\cos s\end{pmatrix}, \quad T(s)=\operatorname{diag}(R(s),R(s)), \quad X(v)=\operatorname{diag}\!\left(\sqrt v,(\sqrt v)^{-1}, (\sqrt v)^{-1},\sqrt v\right).R(s)=(coss−sins​sinscoss​),T(s)=diag(R(s),R(s)),X(v)=diag(v​,(v​)−1,(v​)−1,v​).

This declaration defines the real 4×44\times44×4 matrix, in the order qA,qB,pA,pBq_A,q_B,p_A,p_BqA​,qB​,pA​,pB​, by

typeIICM⁡(gA,gB,t,x)=[T(−t)X(x−1)] diag⁡(gA,gB,gA,gB) [T(−t)X(x−1)]T.\operatorname{typeIICM}(g_A,g_B,t,x) =[T(-t)X(x^{-1})]\, \operatorname{diag}(g_A,g_B,g_A,g_B)\, [T(-t)X(x^{-1})]^{\mathsf T}.typeIICM(gA​,gB​,t,x)=[T(−t)X(x−1)]diag(gA​,gB​,gA​,gB​)[T(−t)X(x−1)]T.

There are no positivity or physicality assumptions on the parameters. Reciprocals of zero are zero and real square roots of nonpositive inputs are zero; consequently the defined matrix is zero when x≤0x\leq0x≤0.

QAlgorithms.Papers.Independent.Intakenext100.P46.weylDisplacedDensity

Let A,BA,BA,B be any finite types, possibly empty, and let σ\sigmaσ be a state indexed by L=NA×NBL=\mathbb N^A\times\mathbb N^BL=NA×NB: its coefficients admit an expansion σuv=∑nan(u)bn(v)‾\sigma_{uv}=\sum_n a_n(u)\overline{b_n(v)}σuv​=∑n​an​(u)bn​(v)​ with an,bn∈ℓ2(L)a_n,b_n\in\ell^2(L)an​,bn​∈ℓ2(L) and ∑n∥an∥∥bn∥<∞\sum_n\|a_n\|\|b_n\|<\infty∑n​∥an​∥∥bn​∥<∞, every finite quadratic form ∑u,v∈Fxu‾σuvxv\sum_{u,v\in F}\overline{x_u}\sigma_{uv}x_v∑u,v∈F​xu​​σuv​xv​ is real and nonnegative, and its real diagonal entries are summable with sum 111. For any real vector zzz indexed by (A⊔B)⊔(A⊔B)(A\sqcup B)\sqcup(A\sqcup B)(A⊔B)⊔(A⊔B), put I=A⊔BI=A\sqcup BI=A⊔B, Ω=(0I−I0)\Omega=\left(\begin{smallmatrix}0&I\\-I&0\end{smallmatrix}\right)Ω=(0−I​I0​), and ξ=Ωz\xi=\Omega zξ=Ωz. Define

d(ζ)rs=e−∣ζ∣2/2r! s!∑h=0min⁡(r,s)ζ r−h(−ζ‾) s−hh! (r−h)! (s−h)!.d(\zeta)_{rs} =e^{-|\zeta|^2/2}\sqrt{r!\,s!} \sum_{h=0}^{\min(r,s)} \frac{\zeta^{\,r-h}(-\overline\zeta)^{\,s-h}} {h!\,(r-h)!\,(s-h)!}.d(ζ)rs​=e−∣ζ∣2/2r!s!​h=0∑min(r,s)​h!(r−h)!(s−h)!ζr−h(−ζ​)s−h​.

Let W(ξ)W(\xi)W(ξ) be the kernel on LLL whose (u,v)(u,v)(u,v)-entry is the product, over modes in AAA and modes in BBB, of d(−(ξqj+iξpj)/2)ujvjd(-(\xi_{q_j}+i\xi_{p_j})/\sqrt2)_{u_jv_j}d(−(ξqj​​+iξpj​​)/2​)uj​vj​​, using the corresponding component of uuu and vvv on each subsystem. The declaration defines the matrix

(weylDisplacedDensity⁡(σ,z))uv=∑a∈LW(Ωz)ua(∑b∈LσabW(Ωz)vb‾).\bigl(\operatorname{weylDisplacedDensity}(\sigma,z)\bigr)_{uv} =\sum_{a\in L}W(\Omega z)_{ua} \left(\sum_{b\in L}\sigma_{ab}\overline{W(\Omega z)_{vb}}\right).(weylDisplacedDensity(σ,z))uv​=a∈L∑​W(Ωz)ua​(b∈L∑​σab​W(Ωz)vb​​).

The displayed nested order of the infinite sums is part of the definition. Its return type is a matrix; this declaration alone supplies no state or separability conclusion for that matrix. Products over empty mode sets are 111.

QAlgorithms.Papers.Independent.Intakenext100.P46.xSqueeze

For every x∈Rx\in\mathbb Rx∈R, label the elements of Fin(1)⊔Fin(1)\mathrm{Fin}(1)\sqcup\mathrm{Fin}(1)Fin(1)⊔Fin(1) by a,ba,ba,b, and order the doubled index set as qa,qb,pa,pbq_a,q_b,p_a,p_bqa​,qb​,pa​,pb​. The declaration defines the real matrix

xSqueeze⁡(x)=diag⁡ ⁣(x,(x)−1,(x)−1,x).\operatorname{xSqueeze}(x) =\operatorname{diag}\!\left(\sqrt x,(\sqrt x)^{-1},(\sqrt x)^{-1},\sqrt x\right).xSqueeze(x)=diag(x​,(x​)−1,(x​)−1,x​).

No positivity assumption on xxx is imposed. The real square root is zero for nonpositive inputs, and 0−1=00^{-1}=00−1=0, so the matrix is zero whenever x≤0x\le0x≤0.

QAlgorithms.Papers.Independent.Intakenext100.P46.ySqueeze

For every real number yyy, this defines the real 4×44\times44×4 matrix

ySqueeze⁡(y)=diag⁡ ⁣(y,y,(y)−1,(y)−1),\operatorname{ySqueeze}(y) =\operatorname{diag}\!\left(\sqrt y,\sqrt y,(\sqrt y)^{-1},(\sqrt y)^{-1}\right),ySqueeze(y)=diag(y​,y​,(y​)−1,(y​)−1),

with two position labels preceding two momentum labels. No positivity assumption is imposed on yyy: the real square root is zero for y≤0y\leq0y≤0, and the reciprocal of zero is defined as zero, so the resulting matrix is zero whenever y≤0y\leq0y≤0.

Human review
  • Endorsed by Shuze Chen · Oct 31, 2026

    Confirmed by the moderator at approval.

  • Endorsed by Gohan · Oct 31, 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