Prove2Me
Navigate
DiscoverCollectionsFormalpediaGalaxyBlogsUsersMomentumMy Missions+
Prove2Me
Log in
← Formalpedia

Constructed separable Gaussian comparison for finite-resource REE

Open
QAlgorithms.Papers.Independent.Intakenext100.P69.separableGaussianComparison

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

prove2quantum

Use physical quadratures (qA,pA,qB,pB)(q_A,p_A,q_B,p_B)(qA​,pA​,qB​,pB​) with ℏ=1\hbar=1ℏ=1 and vacuum covariance I4/2I_4/2I4​/2. Let η>0\eta>0η>0 and r≥0r\ge0r≥0 be finite real numbers. Set

ν=e−2r(η+1)2,D=η+1−e2r∣η−1∣,\nu=\frac{e^{-2r}(\eta+1)}2,\qquad D=\eta+1-e^{2r}|\eta-1|,ν=2e−2r(η+1)​,D=η+1−e2r∣η−1∣,

and assume ν≥∣η−1∣/2\nu\ge |\eta-1|/2ν≥∣η−1∣/2 and D>0D>0D>0. Define

b=−∣η−1∣+ηe2r+e−2r2D,a=2b+(η−1)e−2r2η,c=2b−e−2r2η,csep=(a−1/2)(b−1/2).b=\frac{-|\eta-1|+\eta e^{2r}+e^{-2r}}{2D},\quad a=\frac{2b+(\eta-1)e^{-2r}}{2\eta},\quad c=\frac{2b-e^{-2r}}{2\sqrt\eta},\quad c_{\rm sep}=\sqrt{(a-1/2)(b-1/2)}.b=2D−∣η−1∣+ηe2r+e−2r​,a=2η2b+(η−1)e−2r​,c=2η​2b−e−2r​,csep​=(a−1/2)(b−1/2)​.

Construct the centered Gaussian operators σν\sigma_\nuσν​ and σ~sep\widetilde\sigma_{\rm sep}σsep​ from the respective covariance matrices

V=(aIcZcZbI),V~=(aIcsepZcsepZbI),I=diag⁡(1,1),Z=diag⁡(1,−1).V=\begin{pmatrix}aI&cZ\\cZ&bI\end{pmatrix},\qquad \widetilde V=\begin{pmatrix}aI&c_{\rm sep}Z\\c_{\rm sep}Z&bI\end{pmatrix}, \quad I=\operatorname{diag}(1,1),\quad Z=\operatorname{diag}(1,-1).V=(aIcZ​cZbI​),V=(aIcsep​Z​csep​ZbI​),I=diag(1,1),Z=diag(1,−1).

Both constructed operators are normalized positive trace-class states of finite energy on the full two-mode bosonic Hilbert space, with these physical symmetric covariances. The comparison state is separable across A:BA:BA:B, and

ER(σν):=inf⁡τ mseparableS(σν∥τ)≤S(σν∥σ~sep)=:Ψ(Eη,ν).E_R(\sigma_\nu):=\inf_{\tau\ { m separable}} S(\sigma_\nu\Vert\tau) \le S(\sigma_\nu\Vert\widetilde\sigma_{\rm sep})=: \Psi(\mathcal E_{\eta,\nu}).ER​(σν​):=τ mseparableinf​S(σν​∥τ)≤S(σν​∥σsep​)=:Ψ(Eη,ν​).

Here SSS is extended Umegaki relative entropy in bits and the infimum ranges over every normalized separable state, including non-Gaussian states. This supplies the explicit comparison value used by the paper's finite-resource bounds.

Formalization Note The operators are constructed by their full Weyl characteristics; validity and separability are conclusions. The vacuum extension at covariances without a Gaussian realization is never accepted as a substitute for these conclusions. The accepted characteristic uses doubled covariance, so the adapter supplies 2V2V2V and 2V~2\widetilde V2V. No operational capacity statement is included in this item.

Preamble
import Definitions.Def_QAlgorithms_Papers_Independent_Intakenext100_P69
Formal statement
noncomputable section

open scoped BigOperators ENNReal

namespace QAlgorithms.Papers.Independent.Intakenext100.P69

open QAlgorithms.Papers.Independent.Claimed50.P19
  (Cov DensityKernel GaussianWithCovariance covarianceCharacteristicValue
   multimodeWeylTrace standardFormCovariance vacuumKernel)

open QAlgorithms.Papers.Independent.Dependency50.P33
  (State IsSeparable)

open QAlgorithms.Papers.Independent.Frontier18.P01 (generalizedRelativeEntropy)

attribute [local instance] Classical.propDecidable Classical.decEq

/-- arXiv:1706.06065v3, PDF p.4, (28)--(29), with resource (8)--(11) on p.2.
Both Gaussian states are constructed from their CMs. Normalization, physical
covariance, finite energy, and separability of the comparison are conclusions.
This state comparison does not need or assert the deferred operational converse. -/
theorem separableGaussianComparison
    (eta r : ℝ) (heta : 0 < eta) (hr : 0 ≤ r)
    (hnoise : |eta - 1| / 2 ≤ resourceNoise eta r)
    (hdenominator : 0 < resourceDenominator eta r) :
    CenteredGaussianAt (resourceCovariance eta r) (resourceKernel eta r) ∧
    CenteredGaussianAt (comparisonCovariance eta r) (comparisonKernel eta r) ∧
    IsSeparable (twoModeSplit (comparisonKernel eta r)) ∧
    relativeEntropyEntanglementBits (twoModeSplit (resourceKernel eta r)) ≤
      gaussianComparisonBound eta r := by sorry

end QAlgorithms.Papers.Independent.Intakenext100.P69

end
Source
Finite-resource teleportation stretching for continuous-variable systems, arXiv:1706.06065v3, PDF p. 4, IV, Eqs. (28)–(29); resource Eqs. (8)–(11), p.2
Read-back

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

For all real η,r\eta,rη,r satisfying

η>0,r≥0,∣η−1∣2≤e−2r(η+1)2,d:=η+1−e2r∣η−1∣>0,\eta>0,\qquad r\ge0,\qquad \frac{|\eta-1|}{2}\le\frac{e^{-2r}(\eta+1)}2,\qquad d:=\eta+1-e^{2r}|\eta-1|>0,η>0,r≥0,2∣η−1∣​≤2e−2r(η+1)​,d:=η+1−e2r∣η−1∣>0,

put

b=−∣η−1∣+ηe2r+e−2r2d,a=2b+(η−1)e−2r2η,c=2b−e−2r2η,c∗=(a−12)(b−12),b=\frac{-|\eta-1|+\eta e^{2r}+e^{-2r}}{2d}, \quad a=\frac{2b+(\eta-1)e^{-2r}}{2\eta}, \quad c=\frac{2b-e^{-2r}}{2\sqrt\eta}, \quad c_*=\sqrt{(a-\tfrac12)(b-\tfrac12)},b=2d−∣η−1∣+ηe2r+e−2r​,a=2η2b+(η−1)e−2r​,c=2η​2b−e−2r​,c∗​=(a−21​)(b−21​)​,

and let

V=(a0c00a0−cc0b00−c0b),W=(a0c∗00a0−c∗c∗0b00−c∗0b),V= \begin{pmatrix} a&0&c&0\\0&a&0&-c\\c&0&b&0\\0&-c&0&b \end{pmatrix}, \qquad W= \begin{pmatrix} a&0&c_*&0\\0&a&0&-c_*\\c_*&0&b&0\\0&-c_*&0&b \end{pmatrix},V=​a0c0​0a0−c​c0b0​0−c0b​​,W=​a0c∗​0​0a0−c∗​​c∗​0b0​0−c∗​0b​​,

in coordinate order (0,0),(0,1),(1,0),(1,1)(0,0),(0,1),(1,0),(1,1)(0,0),(0,1),(1,0),(1,1). Identify functions {0,1}→N\{0,1\}\to\mathbb N{0,1}→N with pairs in Q=N2Q=\mathbb N^2Q=N2. For a real 4×44\times44×4 matrix UUU, define the condition C(U,X)\mathcal C(U,X)C(U,X) on a kernel XXX as follows: UUU is symmetric; XXX is a density kernel, meaning that Xpq=∑nun(p)vn(q)‾X_{pq}=\sum_n u_n(p)\overline{v_n(q)}Xpq​=∑n​un​(p)vn​(q)​ for un,vn∈ℓ2(Q;C)u_n,v_n\in\ell^2(Q;\mathbb C)un​,vn​∈ℓ2(Q;C) with ∑n∥un∥∥vn∥\sum_n\|u_n\|\|v_n\|∑n​∥un​∥∥vn​∥ summable, every finite quadratic form ∑p,q∈szp‾Xpqzq\sum_{p,q\in s}\overline{z_p}X_{pq}z_q∑p,q∈s​zp​​Xpq​zq​ is real and nonnegative, and its real diagonal is summable with sum 111; (p0+p1)Re⁡Xpp(p_0+p_1)\operatorname{Re}X_{pp}(p0​+p1​)ReXpp​ is summable; 2U2U2U is positive definite; and 2U+iΩ2U+i\Omega2U+iΩ is positive semidefinite, where Ω=diag⁡(J,J)\Omega=\operatorname{diag}(J,J)Ω=diag(J,J), J=(01−10)J=\left(\begin{smallmatrix}0&1\\-1&0\end{smallmatrix}\right)J=(0−1​10​). Furthermore, there exists m∈R4m\in\mathbb R^4m∈R4 such that for every v∈R4v\in\mathbb R^4v∈R4,

χX(v)=e−12vTUv+i m⋅v,χX(v)=e−12vTUv,χX(v)=∑p∑qXpq∏j=01Dzj(qj,pj),zj=−vj,1+ivj,02,\chi_X(v)=e^{-\frac12v^{\mathsf T}Uv+i\,m\cdot v}, \qquad \chi_X(v)=e^{-\frac12v^{\mathsf T}Uv}, \qquad \chi_X(v)=\sum_p\sum_q X_{pq}\prod_{j=0}^{1}D_{z_j}(q_j,p_j), \quad z_j=\frac{-v_{j,1}+iv_{j,0}}{\sqrt2},χX​(v)=e−21​vTUv+im⋅v,χX​(v)=e−21​vTUv,χX​(v)=p∑​q∑​Xpq​j=0∏1​Dzj​​(qj​,pj​),zj​=2​−vj,1​+ivj,0​​,

where each inner family over qqq and the resulting family of inner sums over ppp is summable, and

Dz(h,k)=e−∣z∣2/2h! k!∑t=0min⁡(h,k)zh−t(−z‾)k−tt! (h−t)! (k−t)!.D_z(h,k)=e^{-|z|^2/2}\sqrt{h!\,k!} \sum_{t=0}^{\min(h,k)} \frac{z^{h-t}(-\overline z)^{k-t}}{t!\,(h-t)!\,(k-t)!}.Dz​(h,k)=e−∣z∣2/2h!k!​t=0∑min(h,k)​t!(h−t)!(k−t)!zh−t(−z)k−t​.

Let K(U)K(U)K(U) be the definition’s chosen XXX satisfying C(U,X)\mathcal C(U,X)C(U,X), if one exists, and otherwise the kernel supported at p=q=(0,0)p=q=(0,0)p=q=(0,0) with value 111. The theorem asserts all four conclusions

C(V,K(V)),C(W,K(W)),K(W) is separable,Ebits(K(V))≤Dbits(K(V),K(W)).\mathcal C(V,K(V)),\qquad \mathcal C(W,K(W)),\qquad K(W)\text{ is separable},\qquad E_{\mathrm{bits}}(K(V)) \le D_{\mathrm{bits}}(K(V),K(W)).C(V,K(V)),C(W,K(W)),K(W) is separable,Ebits​(K(V))≤Dbits​(K(V),K(W)).

Here “separable” means that for every ε>0\varepsilon>0ε>0, K(W)K(W)K(W) has a finite convex-combination approximation YYY of product kernels R⊗SR\otimes SR⊗S with τ(K(W)−Y)<ε\tau(K(W)-Y)<\varepsilonτ(K(W)−Y)<ε, where R,SR,SR,S are density kernels on N\mathbb NN satisfying the density conditions just given, (R⊗S)(i,a),(j,b)=RijSab(R\otimes S)_{(i,a),(j,b)}=R_{ij}S_{ab}(R⊗S)(i,a),(j,b)​=Rij​Sab​, and τ(Z)\tau(Z)τ(Z) is the infimum of [∑n∥un∥∥vn∥]+[\sum_n\|u_n\|\|v_n\|]_+[∑n​∥un​∥∥vn​∥]+​ over all nuclear expansions Zpq=∑nun(p)vn(q)‾Z_{pq}=\sum_n u_n(p)\overline{v_n(q)}Zpq​=∑n​un​(p)vn​(q)​ of the stated kind, with empty infimum +∞+\infty+∞. Also, Ebits(X)E_{\mathrm{bits}}(X)Ebits​(X) is the infimum of Dbits(X,S)D_{\mathrm{bits}}(X,S)Dbits​(X,S) over all separable density kernels SSS on QQQ, and

Dbits(X,Y)=sup⁡s⊆Q finited(Xs,Ys)log⁡2,d(B,C)={[Re⁡Tr⁡(g(B)−Blog⁡C+C−B)]+,ker⁡C⊆ker⁡B,+∞,otherwise.D_{\mathrm{bits}}(X,Y) =\frac{\sup_{s\subseteq Q\text{ finite}}d(X_s,Y_s)}{\log2}, \quad d(B,C)= \begin{cases} [\operatorname{Re}\operatorname{Tr}(g(B)-B\log C+C-B)]_+, &\ker C\subseteq\ker B,\\ +\infty,&\text{otherwise}. \end{cases}Dbits​(X,Y)=log2sups⊆Q finite​d(Xs​,Ys​)​,d(B,C)={[ReTr(g(B)−BlogC+C−B)]+​,+∞,​kerC⊆kerB,otherwise.​

In this formula Xs,YsX_s,Y_sXs​,Ys​ are coordinate restrictions, including the empty one, g(t)=tlog⁡tg(t)=t\log tg(t)=tlogt, matrix functions use real continuous functional calculus, and [t]+=max⁡(t,0)[t]_+=\max(t,0)[t]+​=max(t,0) is embedded in [0,+∞][0,+\infty][0,+∞]. The parameter assumptions include r=0r=0r=0 and equality in the noise inequality; the final inequality is an extended-real inequality and does not separately assert that its right-hand side is finite.

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