Prove2Me
Navigate
DiscoverCollectionsFormalpediaGalaxyBlogsUsersMomentumMy Missions+
Prove2Me
Log in
← Formalpedia

§5: four covariance types and the operational diagonalizations

Open
QAlgorithms.Papers.Independent.Intakenext100.P46.four_types

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

prove2quantum

For a physical standard two-mode covariance let αq=(ac1c1b)\alpha_q=\begin{pmatrix}a&c_1\\c_1&b\end{pmatrix}αq​=(ac1​​c1​b​), αp=(a−c2−c2b)\alpha_p=\begin{pmatrix}a&-c_2\\-c_2&b\end{pmatrix}αp​=(a−c2​​−c2​b​), with c1,c2>0c_1,c_2>0c1​,c2​>0. Select the symmetric chart first if a=ba=ba=b. Otherwise compare A=a/b+b/aA=a/b+b/aA=a/b+b/a and C=c1/c2+c2/c1C=c_1/c_2+c_2/c_1C=c1​/c2​+c2​/c1​: type I has A>CA>CA>C, type II has A<CA<CA<C, and type III has A=CA=CA=C. In type I the local squeeze parameter and the subsequent two-mode squeeze satisfy

x=ac1−bc2ac2−bc1>0,tanh⁡(2r)=−2(ac1−bc2)(ac2−bc1)∣a2−b2∣.x=\sqrt{\frac{ac_1-bc_2}{ac_2-bc_1}}>0,\qquad \tanh(2r)=\frac{-2\sqrt{(ac_1-bc_2)(ac_2-bc_1)}}{|a^2-b^2|}.x=ac2​−bc1​ac1​−bc2​​​>0,tanh(2r)=∣a2−b2∣−2(ac1​−bc2​)(ac2​−bc1​)​​.

In type II they satisfy

x=ac1−bc2bc1−ac2>0,tan⁡(2θ)=2(ac1−bc2)(bc1−ac2)a2−b2sign⁡(bc1−ac2).x=\sqrt{\frac{ac_1-bc_2}{bc_1-ac_2}}>0,\quad \tan(2\theta)=\frac{2\sqrt{(ac_1-bc_2)(bc_1-ac_2)}}{a^2-b^2}\operatorname{sign}(bc_1-ac_2).x=bc1​−ac2​ac1​−bc2​​​>0,tan(2θ)=a2−b22(ac1​−bc2​)(bc1​−ac2​)​​sign(bc1​−ac2​).

Here X(x)X(x)X(x) has diagonal (x,1/x,1/x,x)(\sqrt x,1/\sqrt x,1/\sqrt x,\sqrt x)(x​,1/x​,1/x​,x​); applying R(r)⊕R(−r)R(r)\oplus R(-r)R(r)⊕R(−r) or Θ(θ)⊕Θ(θ)\Theta(\theta)\oplus\Theta(\theta)Θ(θ)⊕Θ(θ) after X(x)X(x)X(x) diagonalizes the covariance. The rotation has rows (cos⁡θ,sin⁡θ)(\cos\theta,\sin\theta)(cosθ,sinθ) and (−sin⁡θ,cos⁡θ)(-\sin\theta,\cos\theta)(−sinθ,cosθ), and RRR has diagonal cosh⁡r\cosh rcoshr and off-diagonal sinh⁡r\sinh rsinhr. In type III, a/b=c1/c2a/b=c_1/c_2a/b=c1​/c2​ or a/b=c2/c1a/b=c_2/c_1a/b=c2​/c1​, and a block symplectic Williamson transformation exists. In the symmetric chart the explicit transformation of Eqs. (12)–(14) applies. This classifies the operational charts without claiming their unprioritized predicates are disjoint. In both generic cases there are further positive local squeeze parameters x′,yx^{\prime},yx′,y for which Y(y)X(x′)Y(y)X(x^{\prime})Y(y)X(x′) converts the resulting diagonal covariance to diag⁡(γA,γB,γA,γB)\operatorname{diag}(\gamma_A,\gamma_B,\gamma_A,\gamma_B)diag(γA​,γB​,γA​,γB​), with γA,γB≥12\gamma_A,\gamma_B\ge\tfrac12γA​,γB​≥21​. Thus the operations give the full normal form, including the final local normalization.

Preamble
import Definitions.Def_QAlgorithms_Papers_Independent_Intakenext100_P46
Formal statement
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)

/-- Chen, PDF pp.8–9, (36): four covariance charts and operational
squeeze/rotation diagonalizations. Ratio predicates overlap at symmetry; the
symmetric chart is selected first. Inverse formulas never run at a=b. -/
theorem four_types (a b c1 c2 : ℝ) (hphysical : IsPhysicalCovariance (standardCM a b c1 c2))
    (hc1 : 0 < c1) (hc2 : 0 < c2) :
    covarianceType a b c1 c2 ∈ ({1, 2, 3, 4} : Set ℕ) ∧
    (a = b →
      let S := symplecticLift (symmetricSq (2 * a) (2 * c1) (2 * c2))
      standardCM a b c1 c2 = S * normalDiagonal
        (symmetricGammas (2 * a) (2 * c1) (2 * c2)) * S.transpose) ∧
    (a ≠ b → c1 / c2 + c2 / c1 < a / b + b / a →
      let x := Real.sqrt ((a * c1 - b * c2) / (a * c2 - b * c1))
      0 < x ∧ ∃ r : ℝ,
        Real.tanh (2 * r) = -2 * Real.sqrt ((a * c1 - b * c2) *
          (a * c2 - b * c1)) / |a ^ 2 - b ^ 2| ∧
        IsDiagonal (congruence (squeezeBoth r * xSqueeze x) (standardCM a b c1 c2)) ∧
        ∃ (xp y : ℝ) (g : Two → ℝ), 0 < xp ∧ 0 < y ∧ (∀ j, 1 / 2 ≤ g j) ∧
          congruence (ySqueeze y * xSqueeze xp * squeezeBoth r * xSqueeze x)
            (standardCM a b c1 c2) = normalDiagonal g) ∧
    (a ≠ b → a / b + b / a < c1 / c2 + c2 / c1 →
      let x := Real.sqrt ((a * c1 - b * c2) / (b * c1 - a * c2))
      0 < x ∧ ∃ t : ℝ,
        Real.tan (2 * t) = 2 * Real.sqrt ((a * c1 - b * c2) *
          (b * c1 - a * c2)) / (a ^ 2 - b ^ 2) * (if 0 < b * c1 - a * c2 then 1 else -1) ∧
        IsDiagonal (congruence (rotateBoth t * xSqueeze x) (standardCM a b c1 c2)) ∧
        ∃ (xp y : ℝ) (g : Two → ℝ), 0 < xp ∧ 0 < y ∧ (∀ j, 1 / 2 ≤ g j) ∧
          congruence (ySqueeze y * xSqueeze xp * rotateBoth t * xSqueeze x)
            (standardCM a b c1 c2) = normalDiagonal g) ∧
    (a ≠ b → a / b + b / a = c1 / c2 + c2 / c1 →
      (a / b = c1 / c2 ∨ a / b = c2 / c1) ∧
      ∃ (Sq : Matrix Two Two ℝ) (g : Two → ℝ),
        IsSymplecticMatrix (symplecticLift Sq) ∧ (∀ j, 1 / 2 ≤ g j) ∧
        standardCM a b c1 c2 = symplecticLift Sq * normalDiagonal g * (symplecticLift Sq).transpose) := by sorry

end QAlgorithms.Papers.Independent.Intakenext100.P46

end
Source
Gaussian relative entropy of entanglement, arXiv:quant-ph/0402109v3, PDF p. 8, §5: four covariance types and the operational diagonalizations
Read-back

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

For all real numbers a,b,c1,c2a,b,c_1,c_2a,b,c1​,c2​, put

α=(ac100c1b0000a−c200−c2b),J=(0I2−I20).\alpha= \begin{pmatrix} a&c_1&0&0\\ c_1&b&0&0\\ 0&0&a&-c_2\\ 0&0&-c_2&b \end{pmatrix}, \qquad J=\begin{pmatrix}0&I_2\\-I_2&0\end{pmatrix}.α=​ac1​00​c1​b00​00a−c2​​00−c2​b​​,J=(0−I2​​I2​0​).

Assume that αT=α\alpha^{\mathsf T}=\alphaαT=α, that the complex matrix α−i2J\alpha-\frac{i}{2}Jα−2i​J is positive semidefinite, and that c1>0c_1>0c1​>0 and c2>0c_2>0c2​>0. Define

H(r)=(cosh⁡rsinh⁡rsinh⁡rcosh⁡r),R(t)=(cos⁡tsin⁡t−sin⁡tcos⁡t),B(r)=diag⁡(H(r),H(−r)),T(t)=diag⁡(R(t),R(t)),X(x)=diag⁡ ⁣(x,(x)−1,(x)−1,x),Y(y)=diag⁡ ⁣(y,y,(y)−1,(y)−1),L(Q)=diag⁡ ⁣(Q,(QT)−1),D(g)=diag⁡(gA,gB,gA,gB)\begin{gathered} H(r)=\begin{pmatrix}\cosh r&\sinh r\\\sinh r&\cosh r\end{pmatrix}, \quad R(t)=\begin{pmatrix}\cos t&\sin t\\-\sin t&\cos t\end{pmatrix},\\ B(r)=\operatorname{diag}(H(r),H(-r)),\qquad T(t)=\operatorname{diag}(R(t),R(t)),\\ X(x)=\operatorname{diag}\!\left(\sqrt x,(\sqrt x)^{-1}, (\sqrt x)^{-1},\sqrt x\right),\\ Y(y)=\operatorname{diag}\!\left(\sqrt y,\sqrt y, (\sqrt y)^{-1},(\sqrt y)^{-1}\right),\\ L(Q)=\operatorname{diag}\!\left(Q,(Q^{\mathsf T})^{-1}\right), \qquad D(g)=\operatorname{diag}(g_A,g_B,g_A,g_B) \end{gathered}H(r)=(coshrsinhr​sinhrcoshr​),R(t)=(cost−sint​sintcost​),B(r)=diag(H(r),H(−r)),T(t)=diag(R(t),R(t)),X(x)=diag(x​,(x​)−1,(x​)−1,x​),Y(y)=diag(y​,y​,(y​)−1,(y​)−1),L(Q)=diag(Q,(QT)−1),D(g)=diag(gA​,gB​,gA​,gB​)​

for a real 2×22\times22×2 matrix QQQ and a function ggg on the two mode labels A,BA,BA,B. The theorem asserts all of the following conjunction: the natural number obtained by choosing 444 when a=ba=ba=b, otherwise 111 when c1/c2+c2/c1<a/b+b/ac_1/c_2+c_2/c_1<a/b+b/ac1​/c2​+c2​/c1​<a/b+b/a, otherwise 222 when the reverse strict inequality holds, and otherwise 333, belongs to {1,2,3,4}\{1,2,3,4\}{1,2,3,4}; if a=ba=ba=b, then, with

Q=12(111−1)diag⁡ ⁣((2a+2c12a−2c2)1/4,(2a−2c12a+2c2)1/4),gA=(2a+2c1)(2a−2c2)2,gB=(2a−2c1)(2a+2c2)2,\begin{gathered} Q=\frac1{\sqrt2} \begin{pmatrix}1&1\\1&-1\end{pmatrix} \operatorname{diag}\!\left( \left(\frac{2a+2c_1}{2a-2c_2}\right)^{1/4}, \left(\frac{2a-2c_1}{2a+2c_2}\right)^{1/4} \right),\\ g_A=\frac{\sqrt{(2a+2c_1)(2a-2c_2)}}2,\qquad g_B=\frac{\sqrt{(2a-2c_1)(2a+2c_2)}}2, \end{gathered}Q=2​1​(11​1−1​)diag((2a−2c2​2a+2c1​​)1/4,(2a+2c2​2a−2c1​​)1/4),gA​=2(2a+2c1​)(2a−2c2​)​​,gB​=2(2a−2c1​)(2a+2c2​)​​,​

one has α=L(Q)D(g)L(Q)T\alpha=L(Q)D(g)L(Q)^{\mathsf T}α=L(Q)D(g)L(Q)T; if a≠ba\ne ba=b and c1/c2+c2/c1<a/b+b/ac_1/c_2+c_2/c_1<a/b+b/ac1​/c2​+c2​/c1​<a/b+b/a, then

x=ac1−bc2ac2−bc1>0x=\sqrt{\frac{ac_1-bc_2}{ac_2-bc_1}}>0x=ac2​−bc1​ac1​−bc2​​​>0

and there exists r∈Rr\in\mathbb Rr∈R such that

tanh⁡(2r)=−2(ac1−bc2)(ac2−bc1)∣a2−b2∣,\tanh(2r)= -\frac{2\sqrt{(ac_1-bc_2)(ac_2-bc_1)}}{|a^2-b^2|},tanh(2r)=−∣a2−b2∣2(ac1​−bc2​)(ac2​−bc1​)​​,

the matrix B(r)X(x)α[B(r)X(x)]TB(r)X(x)\alpha[B(r)X(x)]^{\mathsf T}B(r)X(x)α[B(r)X(x)]T has every off-diagonal entry zero, and there exist real x′>0,y>0x'>0,y>0x′>0,y>0 and a function g:{A,B}→Rg:\{A,B\}\to\mathbb Rg:{A,B}→R, with gj≥12g_j\geq\frac12gj​≥21​ for each jjj, such that

[Y(y)X(x′)B(r)X(x)]α[Y(y)X(x′)B(r)X(x)]T=D(g);[Y(y)X(x')B(r)X(x)]\alpha [Y(y)X(x')B(r)X(x)]^{\mathsf T}=D(g);[Y(y)X(x′)B(r)X(x)]α[Y(y)X(x′)B(r)X(x)]T=D(g);

if a≠ba\ne ba=b and a/b+b/a<c1/c2+c2/c1a/b+b/a<c_1/c_2+c_2/c_1a/b+b/a<c1​/c2​+c2​/c1​, then

x=ac1−bc2bc1−ac2>0x=\sqrt{\frac{ac_1-bc_2}{bc_1-ac_2}}>0x=bc1​−ac2​ac1​−bc2​​​>0

and there exists t∈Rt\in\mathbb Rt∈R such that

tan⁡(2t)=2(ac1−bc2)(bc1−ac2)a2−b2{1,bc1−ac2>0,−1,bc1−ac2≤0,\tan(2t)= \frac{2\sqrt{(ac_1-bc_2)(bc_1-ac_2)}}{a^2-b^2} \begin{cases}1,&bc_1-ac_2>0,\\-1,&bc_1-ac_2\leq0,\end{cases}tan(2t)=a2−b22(ac1​−bc2​)(bc1​−ac2​)​​{1,−1,​bc1​−ac2​>0,bc1​−ac2​≤0,​

the matrix T(t)X(x)α[T(t)X(x)]TT(t)X(x)\alpha[T(t)X(x)]^{\mathsf T}T(t)X(x)α[T(t)X(x)]T has every off-diagonal entry zero, and there exist real x′>0,y>0x'>0,y>0x′>0,y>0 and g:{A,B}→Rg:\{A,B\}\to\mathbb Rg:{A,B}→R, with every gj≥12g_j\geq\frac12gj​≥21​, such that

[Y(y)X(x′)T(t)X(x)]α[Y(y)X(x′)T(t)X(x)]T=D(g);[Y(y)X(x')T(t)X(x)]\alpha [Y(y)X(x')T(t)X(x)]^{\mathsf T}=D(g);[Y(y)X(x′)T(t)X(x)]α[Y(y)X(x′)T(t)X(x)]T=D(g);

and if a≠ba\ne ba=b and a/b+b/a=c1/c2+c2/c1a/b+b/a=c_1/c_2+c_2/c_1a/b+b/a=c1​/c2​+c2​/c1​, then a/b=c1/c2a/b=c_1/c_2a/b=c1​/c2​ or a/b=c2/c1a/b=c_2/c_1a/b=c2​/c1​, and there exist a real 2×22\times22×2 matrix QQQ and g:{A,B}→Rg:\{A,B\}\to\mathbb Rg:{A,B}→R such that L(Q)L(Q)L(Q) is invertible, L(Q)JL(Q)T=JL(Q)JL(Q)^{\mathsf T}=JL(Q)JL(Q)T=J, every gj≥12g_j\geq\frac12gj​≥21​, and α=L(Q)D(g)L(Q)T\alpha=L(Q)D(g)L(Q)^{\mathsf T}α=L(Q)D(g)L(Q)T. These are implications under their respective conditions; scalar division and matrix inversion are total, with scalar division by zero equal to zero and singular matrix inverse equal to zero, and the displayed fourth powers use the total real power function.

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