Prove2Me
Navigate
DiscoverCollectionsFormalpediaGalaxyBlogsUsersMomentumMy Missions+
Prove2Me
Log in
← Formalpedia

§5, Eq. (41): symmetric Gaussian minimizer and two-variable GREE

Open
QAlgorithms.Papers.Independent.Intakenext100.P46.symmetric_gree

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

prove2quantum

Let ρ\rhoρ be an entangled centered symmetric two-mode Gaussian density with covariance blocks 12(mkqkqm)\tfrac12\begin{pmatrix}m&k_q\\k_q&m\end{pmatrix}21​(mkq​​kq​m​) and 12(m−kp−kpm)\tfrac12\begin{pmatrix}m&-k_p\\-k_p&m\end{pmatrix}21​(m−kp​​−kp​m​) in the sign chart kq,kp≥0k_q,k_p\ge0kq​,kp​≥0. Its entropy H(ρ)H(\rho)H(ρ) is finite, even when it has pure modes. For normal eigenvalues γA,γB≥1/2\gamma_A,\gamma_B\ge1/2γA​,γB​≥1/2, put s1=[4γA2(4γB2+1)/(4γA2+1)]1/4s_1=[4\gamma_A^2(4\gamma_B^2+1)/(4\gamma_A^2+1)]^{1/4}s1​=[4γA2​(4γB2​+1)/(4γA2​+1)]1/4, s2=[(4γB2+1)/(4γB2(4γA2+1))]1/4s_2=[(4\gamma_B^2+1)/(4\gamma_B^2(4\gamma_A^2+1))]^{1/4}s2​=[(4γB2​+1)/(4γB2​(4γA2​+1))]1/4 and Q=2−1/2(s1s2s1−s2)Q=2^{-1/2}\begin{pmatrix}s_1&s_2\\s_1&-s_2\end{pmatrix}Q=2−1/2(s1​s1​​s2​−s2​​). Construct the border covariance (Q⊕Q−T)diag⁡(γA,γB,γA,γB)(Q⊕Q−T)T(Q\oplus Q^{-T})\operatorname{diag}(\gamma_A,\gamma_B,\gamma_A,\gamma_B)(Q\oplus Q^{-T})^T(Q⊕Q−T)diag(γA​,γB​,γA​,γB​)(Q⊕Q−T)T, then apply Y(y)−1Y(y)^{-1}Y(y)−1 by covariance congruence, where Y(y)=diag⁡(y,y,y−1/2,y−1/2)Y(y)=\operatorname{diag}(\sqrt y,\sqrt y,y^{-1/2},y^{-1/2})Y(y)=diag(y​,y​,y−1/2,y−1/2) and y>0y>0y>0. Let Vρ(γA,γB)V_\rho(\gamma_A,\gamma_B)Vρ​(γA​,γB​) be the infimum of actual operator divergences to those densities as yyy varies. There exist γA,γB≥1/2\gamma_A,\gamma_B\ge1/2γA​,γB​≥1/2, y>0y>0y>0 and an actual symmetric Gaussian border density attaining

EGR(ρ)=D(ρ∥σ)=Vρ(γA,γB)=min⁡a,b≥1/2Vρ(a,b).E_{GR}(\rho)=D(\rho\Vert\sigma)=V_\rho(\gamma_A,\gamma_B)=\min_{a,b\ge1/2}V_\rho(a,b).EGR​(ρ)=D(ρ∥σ)=Vρ​(γA​,γB​)=a,b≥1/2min​Vρ​(a,b).

For the strict chart put u=ln⁡((2γA+1)/(2γA−1))u=\ln((2\gamma_A+1)/(2\gamma_A-1))u=ln((2γA​+1)/(2γA​−1)) and v=ln⁡((2γB+1)/(2γB−1))v=\ln((2\gamma_B+1)/(2\gamma_B-1))v=ln((2γB​+1)/(2γB​−1)). Then

Vρ=−H(ρ)−ln⁡(2sinh⁡(u/2))−ln⁡(2sinh⁡(v/2))+12(m+kq)(m−kp)u2+(m−kq)(m+kp)v2+(m−kq)(m−kp)uvcoth⁡(u/2)coth⁡(v/2)+(m+kq)(m+kp)uvtanh⁡(u/2)tanh⁡(v/2).V_\rho=-H(\rho)-\ln(2\sinh(u/2))-\ln(2\sinh(v/2))+\tfrac12\sqrt{(m+k_q)(m-k_p)u^2+(m-k_q)(m+k_p)v^2+(m-k_q)(m-k_p)uv\coth(u/2)\coth(v/2)+(m+k_q)(m+k_p)uv\tanh(u/2)\tanh(v/2)}.Vρ​=−H(ρ)−ln(2sinh(u/2))−ln(2sinh(v/2))+21​(m+kq​)(m−kp​)u2+(m−kq​)(m+kp​)v2+(m−kq​)(m−kp​)uvcoth(u/2)coth(v/2)+(m+kq​)(m+kp​)uvtanh(u/2)tanh(v/2)​.

Here EGRE_{GR}EGR​ is the infimum of the actual Umegaki divergence DDD over all separable Gaussian densities on the same cut, with trace-norm closed convex-hull separability. This gives the two-variable minimization. At singular parameters VρV_\rhoVρ​ is defined through the constructed density and independent Umegaki divergence; no unsupported real logarithmic endpoint or finite optimizer is assumed. The common squeeze preserves exchange symmetry and must be retained when kq≠kpk_q\ne k_pkq​=kp​.

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.9,11–12, (41): a symmetric Gaussian border attains GREE;
the resulting optimization has two normal-mode parameters. Endpoints are
operator-defined in symmetricValue, and its literal finite chart is (41).
Pure rho modes are allowed; no faithful-optimizer assumption is inserted. The
common local squeezing is retained when kq and kp differ. Printed on p.9:
symmetric states are called type (iii). Changed: use exchange symmetry, the
a=b/type-(iv) chart. Why: p.8 defines symmetry as type (iv), and the displayed
Eq. (12) and Eq. (41) use that covariance. -/
theorem symmetric_gree (rho : State (Occupation (Fin 1) × Occupation (Fin 1)))
    (m kq kp : ℝ) (hrho : HasGaussian rho 0 (symmetricCM m kq kp))
    (hkq : 0 ≤ kq) (hkp : 0 ≤ kp)
    (hentangled : ¬ IsSeparable rho.coefficients) :
    densityEntropy rho < ⊤ ∧
    (∃ (gA gB y : ℝ) (sigma : State (Occupation (Fin 1) × Occupation (Fin 1))),
      1 / 2 ≤ gA ∧ 1 / 2 ≤ gB ∧ 0 < y ∧
      HasGaussian sigma 0 (symmetricCandidateCM gA gB y) ∧ GaussianBorder sigma ∧
      operatorRelativeEntropy rho sigma = gree rho ∧ symmetricValue rho gA gB = gree rho ∧
      ∀ a b : ℝ, 1 / 2 ≤ a → 1 / 2 ≤ b →
        symmetricValue rho gA gB ≤ symmetricValue rho a b) ∧
    (∀ gA gB : ℝ, 1 / 2 < gA → 1 / 2 < gB →
      symmetricValue rho gA gB = ENNReal.ofReal (-(densityEntropy rho).toReal +
        symmetricObjective m kq kp (modeEM gA) (modeEM gB))) := by sorry

end QAlgorithms.Papers.Independent.Intakenext100.P46

end
Source
Gaussian relative entropy of entanglement, arXiv:quant-ph/0402109v3, PDF p. 11, §5, Eq. (41): symmetric Gaussian minimizer and two-variable GREE
Read-back

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

Let ρ\rhoρ be a state on the label set N×N\mathbb N\times\mathbb NN×N, and let m,kq,kp∈Rm,k_q,k_p\in\mathbb Rm,kq​,kp​∈R. Here, for any label set LLL, a state is a complex matrix RRR admitting sequences ar,br∈ℓ2(L;C)a_r,b_r\in\ell^2(L;\mathbb C)ar​,br​∈ℓ2(L;C) such that ∑r∥ar∥∥br∥<∞\sum_r\|a_r\|\|b_r\|<\infty∑r​∥ar​∥∥br​∥<∞ and Rij=∑rar(i)br(j)‾R_{ij}=\sum_r a_r(i)\overline{b_r(j)}Rij​=∑r​ar​(i)br​(j)​, whose quadratic form on every finitely supported vector is real and nonnegative, and whose real diagonal is summable with sum 111. For any matrix ZZZ, write

∥Z∥nuc=inf⁡(ar,br)ofReal⁡ ⁣(∑r∥ar∥∥br∥),\|Z\|_{\mathrm{nuc}} =\inf_{(a_r,b_r)} \operatorname{ofReal}\!\left(\sum_r\|a_r\|\|b_r\|\right),∥Z∥nuc​=(ar​,br​)inf​ofReal(r∑​∥ar​∥∥br​∥),

where the infimum ranges over those expansions of ZZZ; this takes values in [0,∞][0,\infty][0,∞], with an empty infimum equal to ∞\infty∞, and ofReal⁡(x)=max⁡(x,0)\operatorname{ofReal}(x)=\max(x,0)ofReal(x)=max(x,0). A bipartite matrix is called separable here if, for every ϵ>0\epsilon>0ϵ>0, it is within ϵ\epsilonϵ in this quantity of a finite convex combination of matrices A⊗BA\otimes BA⊗B, where AAA and BBB are states on N\mathbb NN. To specify the Gaussian hypothesis, order the four real coordinates as (qA,qB,pA,pB)(q_A,q_B,p_A,p_B)(qA​,qB​,pA​,pB​), put

J=(0I2−I20),α(m,kq,kp)=12(mkq00kqm0000m−kp00−kpm),J=\begin{pmatrix}0&I_2\\-I_2&0\end{pmatrix}, \qquad \alpha(m,k_q,k_p)=\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},J=(0−I2​​I2​0​),α(m,kq​,kp​)=21​​mkq​00​kq​m00​00m−kp​​00−kp​m​​,

and define the displacement coefficients, for r,s∈Nr,s\in\mathbb Nr,s∈N, by

Dz(r,s)=e−∣z∣2/2r! s!∑h=0min⁡(r,s)zr−h(−z‾)s−hh! (r−h)! (s−h)!.D_z(r,s)=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)!}.Dz​(r,s)=e−∣z∣2/2r!s!​h=0∑min(r,s)​h!(r−h)!(s−h)!zr−h(−z)s−h​.

For ξ∈R4\xi\in\mathbb R^4ξ∈R4, set zA=−(qA+ipA)/2z_A=-(q_A+ip_A)/\sqrt2zA​=−(qA​+ipA​)/2​, zB=−(qB+ipB)/2z_B=-(q_B+ip_B)/\sqrt2zB​=−(qB​+ipB​)/2​, and

Wξ(u,v)=DzA(uA,vA)DzB(uB,vB).W_\xi(u,v)=D_{z_A}(u_A,v_A)D_{z_B}(u_B,v_B).Wξ​(u,v)=DzA​​(uA​,vA​)DzB​​(uB​,vB​).

Write G(R,α)G(R,\alpha)G(R,α) for the following conditions: αT=α\alpha^{\mathsf T}=\alphaαT=α, α−i2J\alpha-\tfrac i2Jα−2i​J is positive semidefinite, and, for every ξ\xiξ, each inner series below and the outer series 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).

Assume G(ρ,α(m,kq,kp))G(\rho,\alpha(m,k_q,k_p))G(ρ,α(m,kq​,kp​)), kq≥0k_q\ge0kq​≥0, kp≥0k_p\ge0kp​≥0, and that ρ\rhoρ is not separable. Define R^\widehat RR to be a chosen bounded operator on ℓ2(L;C)\ell^2(L;\mathbb C)ℓ2(L;C) with matrix RRR when one exists, and zero otherwise. With eie_iei​ the coordinate vectors and functional calculus denoted by f(A)f(A)f(A), define

S(R)=∑iofReal⁡(Re⁡⟨ei,(−xlog⁡x)(R^)ei⟩).S(R)=\sum_i\operatorname{ofReal} \bigl(\operatorname{Re}\langle e_i,(-x\log x)(\widehat R)e_i\rangle\bigr).S(R)=i∑​ofReal(Re⟨ei​,(−xlogx)(R)ei​⟩).

For finite complex matrices A,BA,BA,B, define

d(A,B)={ofReal⁡Re⁡Tr⁡((xlog⁡x)(A)−Alog⁡(B)−A+B),ker⁡B⊆ker⁡A,∞,otherwise,d(A,B)= \begin{cases} \operatorname{ofReal}\operatorname{Re}\operatorname{Tr} \bigl((x\log x)(A)-A\log(B)-A+B\bigr), &\ker B\subseteq\ker A,\\ \infty,&\text{otherwise}, \end{cases}d(A,B)={ofRealReTr((xlogx)(A)−Alog(B)−A+B),∞,​kerB⊆kerA,otherwise,​

and define

D(R,T)=sup⁡n∈Nv0,…,vn−1 orthonormald ⁣([⟨vi,R^vj⟩]i,j,[⟨vi,T^vj⟩]i,j),EG(R)=inf⁡T a separable state∃d,β: Gd(T,β)D(R,T),D(R,T)= \sup_{\substack{n\in\mathbb N\\v_0,\ldots,v_{n-1}\ \mathrm{orthonormal}}} d\!\left( [\langle v_i,\widehat Rv_j\rangle]_{i,j}, [\langle v_i,\widehat Tv_j\rangle]_{i,j} \right), \qquad E_G(R)=\inf_{\substack{T\ \mathrm{a\ separable\ state}\\ \exists d,\beta:\ G_d(T,\beta)}}D(R,T),D(R,T)=n∈Nv0​,…,vn−1​ orthonormal​sup​d([⟨vi​,Rvj​⟩]i,j​,[⟨vi​,Tvj​⟩]i,j​),EG​(R)=T a separable state∃d,β: Gd​(T,β)​inf​D(R,T),

where Gd(T,β)G_d(T,\beta)Gd​(T,β) has the same physicality and summability conditions as GGG, and characteristic function exp⁡(−ξTJTβJξ/2+iξTJd)\exp(-\xi^{\mathsf T}J^{\mathsf T}\beta J\xi/2+i\xi^{\mathsf T}Jd)exp(−ξTJTβJξ/2+iξTJd). The supremum defining DDD includes n=0n=0n=0. For real a,ba,ba,b, put

H=(111−1),Qa,b=12H diag⁡ ⁣((4a2(4b2+1)4a2+1)1/4,(4b2+14b2(4a2+1))1/4),La,b=diag⁡(Qa,b,Qa,b−T),Yy=diag⁡(y−1/2I2,y1/2I2)(y>0),Ba,b,y=YyLa,bdiag⁡(a,b,a,b)La,bTYyT,Vρ(a,b)=inf⁡y>0, T a stateG(T,Ba,b,y)D(ρ,T).\begin{gathered} H=\begin{pmatrix}1&1\\1&-1\end{pmatrix},\qquad Q_{a,b}=\frac1{\sqrt2}H\, \operatorname{diag}\!\left( \left(\frac{4a^2(4b^2+1)}{4a^2+1}\right)^{1/4}, \left(\frac{4b^2+1}{4b^2(4a^2+1)}\right)^{1/4} \right),\\ L_{a,b}=\operatorname{diag}(Q_{a,b},Q_{a,b}^{-\mathsf T}), \qquad Y_y=\operatorname{diag}(y^{-1/2}I_2,y^{1/2}I_2)\quad(y>0),\\ B_{a,b,y} =Y_yL_{a,b}\operatorname{diag}(a,b,a,b)L_{a,b}^{\mathsf T}Y_y^{\mathsf T}, \qquad V_\rho(a,b)= \inf_{\substack{y>0,\ T\ \mathrm{a\ state}\\G(T,B_{a,b,y})}}D(\rho,T). \end{gathered}H=(11​1−1​),Qa,b​=2​1​Hdiag((4a2+14a2(4b2+1)​)1/4,(4b2(4a2+1)4b2+1​)1/4),La,b​=diag(Qa,b​,Qa,b−T​),Yy​=diag(y−1/2I2​,y1/2I2​)(y>0),Ba,b,y​=Yy​La,b​diag(a,b,a,b)La,bT​YyT​,Vρ​(a,b)=y>0, T a stateG(T,Ba,b,y​)​inf​D(ρ,T).​

Then S(ρ)<∞S(\rho)<\inftyS(ρ)<∞, and there exist real gA,gB,yg_A,g_B,ygA​,gB​,y and a state σ\sigmaσ such that gA,gB≥12g_A,g_B\ge\tfrac12gA​,gB​≥21​, y>0y>0y>0, G(σ,BgA,gB,y)G(\sigma,B_{g_A,g_B,y})G(σ,BgA​,gB​,y​), σ\sigmaσ is separable, and σ\sigmaσ admits a centered Gaussian covariance β\betaβ such that, for every ϵ>0\epsilon>0ϵ>0, there are a nonseparable state τ\tauτ and covariance β′\beta'β′ with G(τ,β′)G(\tau,\beta')G(τ,β′) and ∣βij′−βij∣<ϵ|\beta'_{ij}-\beta_{ij}|<\epsilon∣βij′​−βij​∣<ϵ for every entry. These witnesses satisfy

D(ρ,σ)=EG(ρ),Vρ(gA,gB)=EG(ρ),Vρ(gA,gB)≤Vρ(a,b)for every a,b≥12.D(\rho,\sigma)=E_G(\rho),\qquad V_\rho(g_A,g_B)=E_G(\rho),\qquad V_\rho(g_A,g_B)\le V_\rho(a,b) \quad\text{for every }a,b\ge\tfrac12.D(ρ,σ)=EG​(ρ),Vρ​(gA​,gB​)=EG​(ρ),Vρ​(gA​,gB​)≤Vρ​(a,b)for every a,b≥21​.

Furthermore, for every gA,gB>12g_A,g_B>\tfrac12gA​,gB​>21​, setting u=log⁡2gA+12gA−1u=\log\frac{2g_A+1}{2g_A-1}u=log2gA​−12gA​+1​ and v=log⁡2gB+12gB−1v=\log\frac{2g_B+1}{2g_B-1}v=log2gB​−12gB​+1​ gives

Vρ(gA,gB)=ofReal⁡(−S(ρ)R−log⁡(2sinh⁡(u/2))−log⁡(2sinh⁡(v/2))+12(m+kq)(m−kp)u2+(m−kq)(m+kp)v2+(m−kq)(m−kp)uvcosh⁡(u/2)cosh⁡(v/2)sinh⁡(u/2)sinh⁡(v/2)+(m+kq)(m+kp)uvtanh⁡(u/2)tanh⁡(v/2)).\begin{aligned} V_\rho(g_A,g_B)=\operatorname{ofReal}\Bigg( &-S(\rho)_{\mathbb R} -\log(2\sinh(u/2))-\log(2\sinh(v/2))\\ &+\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)\cosh(v/2)}{\sinh(u/2)\sinh(v/2)} +(m+k_q)(m+k_p)uv\tanh(u/2)\tanh(v/2)} \Bigg). \end{aligned}Vρ​(gA​,gB​)=ofReal(​−S(ρ)R​−log(2sinh(u/2))−log(2sinh(v/2))+21​(m+kq​)(m−kp​)u2+(m−kq​)(m+kp​)v2+(m−kq​)(m−kp​)uvsinh(u/2)sinh(v/2)cosh(u/2)cosh(v/2)​+(m+kq​)(m+kp​)uvtanh(u/2)tanh(v/2)​).​

Here S(ρ)RS(\rho)_{\mathbb R}S(ρ)R​ is the real conversion of S(ρ)S(\rho)S(ρ), finite by the conclusion. The existence and minimization statements include gA=12g_A=\tfrac12gA​=21​ or gB=12g_B=\tfrac12gB​=21​; the displayed objective formula is asserted only when both inequalities are strict.

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