Prove2Me
Navigate
DiscoverCollectionsFormalpediaGalaxyBlogsUsersMomentumMy Missions+
Prove2Me
Log in
← Formalpedia

Definitions for Finite-resource teleportation stretching for continuous-variable systems

Definition
QAlgorithms_Papers_Independent_Intakenext100_P69

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

prove2quantum

The objects the formalized statements of Finite-resource teleportation stretching for continuous-variable systems (arXiv:1706.06065) are stated in. Each was checked against the paper by a blind read-back and a moderator.

  • CenteredGaussianAt: PDF pp.2,8 [59]: centered normalized trace-class Gaussian density with physical covariance V, finite energy, convergent Weyl traces, and real symmetry. The accepted predicate uses doubled covariance (vacuum I), hence 2 times V.
  • centeredTwoModeGaussianKernel: PDF pp.2,4: construct the unique centered Gaussian density from its full Weyl characteristic, using choice only when it exists. Vacuum is an explicit off-domain extension. The comparison theorem CONCLUDES validity at both constructed covariances, so the fallback cannot discharge a physical claim.
  • resourceDenominator: PDF p.2, (10): D before the overall factor 2 in the denominator of b.
  • resourceNoise: PDF p.2, (11), in physical vacuum variance 1/2 units.
  • resourceB: PDF p.2, (10), restricted to D>0 in the theorem.
  • resourceA: PDF p.2, (9), restricted to eta>0 in the theorem.
  • resourceC: PDF p.2, (9): the positive square root of the intensity gain.
  • resourceCovariance: PDF p.2, (8): (q_A,p_A,q_B,p_B), with off-diagonal c diag(1,-1).
  • comparisonC: PDF p.4, (28): comparison correlation computed from this resource.
  • comparisonCovariance: PDF p.4, (28): keep both diagonal blocks, replace only c by c_sep.
  • resourceKernel: PDF p.2, (8)--(11): the actual centered resource density, not its CM.
  • comparisonKernel: PDF p.4, (28): the actual centered comparison density, not a free candidate.
  • twoModeSplit: PDF pp.2,4: relabel full two-mode Fock coordinates as Alice times Bob. Neither occupation number is truncated.
  • normalStateRelativeEntropyBits: PDF p.3: Umegaki relative entropy in bits on actual normal densities. The accepted extended finite-compression definition retains support failures and divergence as infinity; no difference of infinite entropies is taken.
  • relativeEntropyEntanglementBits: PDF pp.3--4, (23),(29): infimum over ALL normalized separable densities, using their full trace-norm closed convex-hull definition, not only Gaussians.
  • gaussianComparisonBound: PDF p.4, (29): Psi is the divergence to the constructed Gaussian candidate.
  • additiveComparisonBound: PDF pp.2,5, (11), eta=1: r=-ln(xi)/2. Used physically for 0<xi<=1.
  • additiveChainComparisonBound: PDF p.6, (39): minimum of the constructed additive-noise comparison values on all N+1 links. The index type is nonempty even when N=0.
Definition code
import Mathlib
import Definitions.Def_QAlgorithms_Papers_Independent_Claimed50_P19
import Definitions.Def_QAlgorithms_Papers_Independent_Frontier18_P01

/-!
Laurenza--Braunstein--Pirandola, arXiv:1706.06065v3, PDF pp.2--8.
The two statements below are the state comparison (28)--(29) and the algebraic
chain specialization on p.6. Operational capacity/stretching nodes are deferred
in HARD.md/GAPS.json: no finite-dimensional LOCC surrogate is used.
Covariance is in interleaved physical vacuum-I/2 units. Only the adapter to the
accepted characteristic doubles it. All densities have full oscillator support.
No defective Claimed50.P47.HasBipartiteMoments predicate is used.
Compilation and independent moderation are pending the batch handoff.
-/

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

/-- PDF pp.2,8 [59]: centered normalized trace-class Gaussian density with
physical covariance V, finite energy, convergent Weyl traces, and real symmetry.
The accepted predicate uses doubled covariance (vacuum I), hence 2 times V. -/
def CenteredGaussianAt (V : Cov 2) (rho : DensityKernel 2) : Prop :=
  V.transpose = V ∧ GaussianWithCovariance rho ((2 : ℝ) • V) ∧
    ∀ v, multimodeWeylTrace rho v =
      covarianceCharacteristicValue ((2 : ℝ) • V) 0 v

/-- PDF pp.2,4: construct the unique centered Gaussian density from its full
Weyl characteristic, using choice only when it exists. Vacuum is an explicit
off-domain extension. The comparison theorem CONCLUDES validity at both
constructed covariances, so the fallback cannot discharge a physical claim. -/
def centeredTwoModeGaussianKernel (V : Cov 2) : DensityKernel 2 :=
  if h : ∃ rho : DensityKernel 2, CenteredGaussianAt V rho then
    Classical.choose h
  else vacuumKernel 2

/-- PDF p.2, (10): D before the overall factor 2 in the denominator of b. -/
def resourceDenominator (eta r : ℝ) : ℝ :=
  eta + 1 - Real.exp (2 * r) * |eta - 1|

/-- PDF p.2, (11), in physical vacuum variance 1/2 units. -/
def resourceNoise (eta r : ℝ) : ℝ :=
  Real.exp (-2 * r) * (eta + 1) / 2

/-- PDF p.2, (10), restricted to D>0 in the theorem. -/
def resourceB (eta r : ℝ) : ℝ :=
  (-|eta - 1| + eta * Real.exp (2 * r) + Real.exp (-2 * r)) /
    (2 * resourceDenominator eta r)

/-- PDF p.2, (9), restricted to eta>0 in the theorem. -/
def resourceA (eta r : ℝ) : ℝ :=
  (2 * resourceB eta r + (eta - 1) * Real.exp (-2 * r)) / (2 * eta)

/-- PDF p.2, (9): the positive square root of the intensity gain. -/
def resourceC (eta r : ℝ) : ℝ :=
  (2 * resourceB eta r - Real.exp (-2 * r)) / (2 * Real.sqrt eta)

/-- PDF p.2, (8): (q_A,p_A,q_B,p_B), with off-diagonal c diag(1,-1). -/
def resourceCovariance (eta r : ℝ) : Cov 2 :=
  standardFormCovariance (resourceA eta r) (resourceB eta r)
    (resourceC eta r) (-resourceC eta r)

/-- PDF p.4, (28): comparison correlation computed from this resource. -/
def comparisonC (eta r : ℝ) : ℝ :=
  Real.sqrt ((resourceA eta r - 1 / 2) * (resourceB eta r - 1 / 2))

/-- PDF p.4, (28): keep both diagonal blocks, replace only c by c_sep. -/
def comparisonCovariance (eta r : ℝ) : Cov 2 :=
  standardFormCovariance (resourceA eta r) (resourceB eta r)
    (comparisonC eta r) (-comparisonC eta r)

/-- PDF p.2, (8)--(11): the actual centered resource density, not its CM. -/
def resourceKernel (eta r : ℝ) : DensityKernel 2 :=
  centeredTwoModeGaussianKernel (resourceCovariance eta r)

/-- PDF p.4, (28): the actual centered comparison density, not a free candidate. -/
def comparisonKernel (eta r : ℝ) : DensityKernel 2 :=
  centeredTwoModeGaussianKernel (comparisonCovariance eta r)

/-- PDF pp.2,4: relabel full two-mode Fock coordinates as Alice times Bob.
Neither occupation number is truncated. -/
def twoModeSplit (rho : DensityKernel 2) : Matrix (ℕ × ℕ) (ℕ × ℕ) ℂ :=
  fun p q => rho (fun i => if i = 0 then p.1 else p.2)
    (fun i => if i = 0 then q.1 else q.2)

/-- PDF p.3: Umegaki relative entropy in bits on actual normal densities.
The accepted extended finite-compression definition retains support failures
and divergence as infinity; no difference of infinite entropies is taken. -/
def normalStateRelativeEntropyBits {I : Type} (rho sigma : Matrix I I ℂ) : ℝ≥0∞ :=
  generalizedRelativeEntropy rho sigma / ENNReal.ofReal (Real.log 2)

/-- PDF pp.3--4, (23),(29): infimum over ALL normalized separable densities,
using their full trace-norm closed convex-hull definition, not only Gaussians. -/
def relativeEntropyEntanglementBits (rho : Matrix (ℕ × ℕ) (ℕ × ℕ) ℂ) : ℝ≥0∞ :=
  ⨅ sigma : {sigma : State (ℕ × ℕ) // IsSeparable sigma.coefficients},
    normalStateRelativeEntropyBits rho sigma.val.coefficients

/-- PDF p.4, (29): Psi is the divergence to the constructed Gaussian candidate. -/
def gaussianComparisonBound (eta r : ℝ) : ℝ≥0∞ :=
  normalStateRelativeEntropyBits (twoModeSplit (resourceKernel eta r))
    (twoModeSplit (comparisonKernel eta r))

/-- PDF pp.2,5, (11), eta=1: r=-ln(xi)/2. Used physically for 0<xi<=1. -/
def additiveComparisonBound (xi : ℝ) : ℝ≥0∞ :=
  gaussianComparisonBound 1 (-Real.log xi / 2)

/-- PDF p.6, (39): minimum of the constructed additive-noise comparison values
on all N+1 links. The index type is nonempty even when N=0. -/
def additiveChainComparisonBound (N : ℕ) (noise : Fin (N + 1) → ℝ) : ℝ≥0∞ :=
  ⨅ i : Fin (N + 1), additiveComparisonBound (noise i)

/-- Sanity check for p.2, (8): q_A--q_B has sign +c. -/
example (eta r : ℝ) :
    resourceCovariance eta r (0, 0) (1, 0) = resourceC eta r := by
  simp [resourceCovariance, standardFormCovariance]

/-- Sanity check for p.2, (8): p_A--p_B has sign -c. -/
example (eta r : ℝ) :
    resourceCovariance eta r (0, 1) (1, 1) = -resourceC eta r := by
  simp [resourceCovariance, standardFormCovariance]

/-- Sanity check: the fallback is explicit and confined to absent Gaussians. -/
example (V : Cov 2) (h : ¬ ∃ rho, CenteredGaussianAt V rho) :
    centeredTwoModeGaussianKernel V = vacuumKernel 2 := by
  simp [centeredTwoModeGaussianKernel, h]

/-- Sanity check: at eta=1,r=0, the resource is prescribed the vacuum CM I/2. -/
example : resourceA 1 0 = 1 / 2 ∧ resourceB 1 0 = 1 / 2 ∧
    resourceC 1 0 = 0 := by
  norm_num [resourceA, resourceB, resourceC, resourceDenominator]

/-- Sanity check: at xi=1 the additive parametrization has r=0. -/
example : additiveComparisonBound 1 = gaussianComparisonBound 1 0 := by
  simp [additiveComparisonBound]

end QAlgorithms.Papers.Independent.Intakenext100.P69
Source
Finite-resource teleportation stretching for continuous-variable systems, arXiv:1706.06065v3, https://arxiv.org/abs/1706.06065v3
Read-back

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

QAlgorithms.Papers.Independent.Intakenext100.P69.CenteredGaussianAt

For any real matrix VVV indexed by {0,1}×{0,1}\{0,1\}\times\{0,1\}{0,1}×{0,1} and any complex kernel ρ=(ρij)i,j∈N2\rho=(\rho_{ij})_{i,j\in\mathbb N^2}ρ=(ρij​)i,j∈N2​, this predicate requires the following conjunction. First, VT=VV^{\mathsf T}=VVT=V. Second, ρ\rhoρ has a representation ρij=∑n=0∞un(i)wn(j)‾\rho_{ij}=\sum_{n=0}^{\infty}u_n(i)\overline{w_n(j)}ρij​=∑n=0∞​un​(i)wn​(j)​, where un,wn∈ℓ2(N2;C)u_n,w_n\in\ell^2(\mathbb N^2;\mathbb C)un​,wn​∈ℓ2(N2;C) and ∑n∥un∥∥wn∥\sum_n\|u_n\|\|w_n\|∑n​∥un​∥∥wn​∥ is summable; for every finite s⊆N2s\subseteq\mathbb N^2s⊆N2 and every z:N2→Cz:\mathbb N^2\to\mathbb Cz:N2→C, the number ∑i,j∈szi‾ρijzj\sum_{i,j\in s}\overline{z_i}\rho_{ij}z_j∑i,j∈s​zi​​ρij​zj​ has zero imaginary part and nonnegative real part; the real diagonal entries are summable with ∑iRe⁡ρii=1\sum_i\operatorname{Re}\rho_{ii}=1∑i​Reρii​=1; 2V2V2V is positive definite; 2V+iΩ2V+i\Omega2V+iΩ is positive semidefinite, where Ω=diag⁡(J,J)\Omega=\operatorname{diag}(J,J)Ω=diag(J,J) in the 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), with J=(01−10)J=\begin{pmatrix}0&1\\-1&0\end{pmatrix}J=(0−1​10​); and the real family ((i0+i1)Re⁡ρii)i∈N2\bigl((i_0+i_1)\operatorname{Re}\rho_{ii}\bigr)_{i\in\mathbb N^2}((i0​+i1​)Reρii​)i∈N2​ is summable. To state the remaining requirements, for every real vector v=(vk,a)k,a∈{0,1}v=(v_{k,a})_{k,a\in\{0,1\}}v=(vk,a​)k,a∈{0,1}​, put zk=(−vk,1+ivk,0)/2z_k=(-v_{k,1}+iv_{k,0})/\sqrt2zk​=(−vk,1​+ivk,0​)/2​ and

Dz(a,b)=e−∣z∣2/2a! b!∑h=0min⁡(a,b)za−h(−z‾)b−hh! (a−h)! (b−h)!,Tij(v)=ρij∏k=01Dzk(jk,ik).D_z(a,b)=e^{-|z|^2/2}\sqrt{a!\,b!} \sum_{h=0}^{\min(a,b)} \frac{z^{a-h}(-\overline z)^{b-h}}{h!\,(a-h)!\,(b-h)!}, \qquad T_{ij}(v)=\rho_{ij}\prod_{k=0}^{1}D_{z_k}(j_k,i_k).Dz​(a,b)=e−∣z∣2/2a!b!​h=0∑min(a,b)​h!(a−h)!(b−h)!za−h(−z)b−h​,Tij​(v)=ρij​k=0∏1​Dzk​​(jk​,ik​).

There must exist a real vector mmm indexed by {0,1}×{0,1}\{0,1\}\times\{0,1\}{0,1}×{0,1} such that, for every vvv, each family (Tij(v))j(T_{ij}(v))_j(Tij​(v))j​ is summable, the family (∑jTij(v))i(\sum_jT_{ij}(v))_i(∑j​Tij​(v))i​ is summable, and

∑i∑jTij(v)=exp⁡ ⁣(−14∑a,bva(2V)abvb+i∑amava).\sum_i\sum_jT_{ij}(v) =\exp\!\left(-\frac14\sum_{a,b}v_a(2V)_{ab}v_b+i\sum_a m_av_a\right).i∑​j∑​Tij​(v)=exp​−41​a,b∑​va​(2V)ab​vb​+ia∑​ma​va​​.

Finally, independently of that existential choice of mmm, the same iterated sum must satisfy, for every vvv,

∑i∑jTij(v)=exp⁡ ⁣(−14∑a,bva(2V)abvb).\sum_i\sum_jT_{ij}(v) =\exp\!\left(-\frac14\sum_{a,b}v_a(2V)_{ab}v_b\right).i∑​j∑​Tij​(v)=exp​−41​a,b∑​va​(2V)ab​vb​​.

QAlgorithms.Papers.Independent.Intakenext100.P69.additiveChainComparisonBound

For every natural number NNN, including N=0N=0N=0, and every function ν:{0,…,N}→R\nu:\{0,\ldots,N\}\to\mathbb Rν:{0,…,N}→R, this defines

CN(ν)=inf⁡0≤i≤NA(νi),C_N(\nu)=\inf_{0\le i\le N}A(\nu_i),CN​(ν)=0≤i≤Ninf​A(νi​),

an infimum in [0,+∞][0,+\infty][0,+∞] over N+1N+1N+1 indices, equivalently their minimum. To specify each value A(x)A(x)A(x) fully, put r=−log⁡(x)/2r=-\log(x)/2r=−log(x)/2 and

a=b=e2r+e−2r4,c=e2r−e−2r4,c∗=(a−12)(b−12),a=b=\frac{e^{2r}+e^{-2r}}4, \qquad c=\frac{e^{2r}-e^{-2r}}4, \qquad c_*=\sqrt{(a-\tfrac12)(b-\tfrac12)},a=b=4e2r+e−2r​,c=4e2r−e−2r​,c∗​=(a−21​)(b−21​)​,

and form, 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),

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​​.

Then A(x)=Dbits(K(V),K(W))A(x)=D_{\mathrm{bits}}(K(V),K(W))A(x)=Dbits​(K(V),K(W)). With functions {0,1}→N\{0,1\}\to\mathbb N{0,1}→N identified with Q=N2Q=\mathbb N^2Q=N2, K(U)K(U)K(U) is the definition’s chosen kernel XXX satisfying all of these requirements, if one exists, and otherwise is the kernel supported at p=q=(0,0)p=q=(0,0)p=q=(0,0) with value 111: UUU is symmetric, 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) and J=(01−10)J=\left(\begin{smallmatrix}0&1\\-1&0\end{smallmatrix}\right)J=(0−1​10​); Xpq=∑nun(p)vn(q)‾X_{pq}=\sum_n u_n(p)\overline{v_n(q)}Xpq​=∑n​un​(p)vn​(q)​ with un,vn∈ℓ2(Q;C)u_n,v_n\in\ell^2(Q;\mathbb C)un​,vn​∈ℓ2(Q;C) and ∑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; the real diagonal is summable with sum 111; and (p0+p1)Re⁡Xpp(p_0+p_1)\operatorname{Re}X_{pp}(p0​+p1​)ReXpp​ is summable. There must also exist m∈R4m\in\mathbb R^4m∈R4 such that for every v∈R4v\in\mathbb R^4v∈R4,

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

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

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​.

Here

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,​

with g(t)=tlog⁡tg(t)=t\log tg(t)=tlogt, real continuous functional calculus for matrix functions, coordinate restrictions Xs,YsX_s,Y_sXs​,Ys​, and [t]+=max⁡(t,0)[t]_+=\max(t,0)[t]+​=max(t,0) embedded in the extended nonnegative reals; the supremum includes s=∅s=\varnothings=∅. No positivity assumption is imposed on any νi\nu_iνi​: log⁡0=0\log0=0log0=0, the logarithm of a negative real equals that of its absolute value, and real square roots of negative arguments are zero. For N=0N=0N=0, the defined infimum has just the single value A(ν0)A(\nu_0)A(ν0​).

QAlgorithms.Papers.Independent.Intakenext100.P69.additiveComparisonBound

For every real ξ\xiξ, this defines the extended nonnegative number obtained from the following construction with η=1\eta=1η=1 and r=−log⁡(ξ)/2r=-\log(\xi)/2r=−log(ξ)/2:

A(ξ)=Dbits(K(Vη,r),K(Wη,r)),A(\xi)=D_{\mathrm{bits}}\bigl(K(V_{\eta,r}),K(W_{\eta,r})\bigr),A(ξ)=Dbits​(K(Vη,r​),K(Wη,r​)),

where

d=η+1−e2r∣η−1∣,b=−∣η−1∣+ηe2r+e−2r2d,a=2b+(η−1)e−2r2η,c=2b−e−2r2η,c∗=(a−12)(b−12),Vη,r=(a0c00a0−cc0b00−c0b),Wη,r=(a0c∗00a0−c∗c∗0b00−c∗0b).\begin{gathered} d=\eta+1-e^{2r}|\eta-1|, \quad 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)},\\ V_{\eta,r}= \begin{pmatrix} a&0&c&0\\0&a&0&-c\\c&0&b&0\\0&-c&0&b \end{pmatrix}, \qquad W_{\eta,r}= \begin{pmatrix} a&0&c_*&0\\0&a&0&-c_*\\c_*&0&b&0\\0&-c_*&0&b \end{pmatrix}. \end{gathered}d=η+1−e2r∣η−1∣,b=2d−∣η−1∣+ηe2r+e−2r​,a=2η2b+(η−1)e−2r​,c=2η​2b−e−2r​,c∗​=(a−21​)(b−21​)​,Vη,r​=​a0c0​0a0−c​c0b0​0−c0b​​,Wη,r​=​a0c∗​0​0a0−c∗​​c∗​0b0​0−c∗​0b​​.​

Coordinates are ordered as (0,0),(0,1),(1,0),(1,1)(0,0),(0,1),(1,0),(1,1)(0,0),(0,1),(1,0),(1,1). Identifying functions {0,1}→N\{0,1\}\to\mathbb N{0,1}→N with pairs Q=N2Q=\mathbb N^2Q=N2, K(U)K(U)K(U) is the definition’s chosen kernel XXX satisfying the conditions below, if one exists, and otherwise is Xpq=1X_{pq}=1Xpq​=1 for p=q=(0,0)p=q=(0,0)p=q=(0,0) and 000 elsewhere. Those conditions are: UUU is symmetric; 2U2U2U is positive definite and 2U+iΩ2U+i\Omega2U+iΩ is positive semidefinite, with 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​); Xpq=∑nun(p)vn(q)‾X_{pq}=\sum_n u_n(p)\overline{v_n(q)}Xpq​=∑n​un​(p)vn​(q)​ for sequences in ℓ2(Q;C)\ell^2(Q;\mathbb C)ℓ2(Q;C) with ∑n∥un∥∥vn∥\sum_n\|u_n\|\|v_n\|∑n​∥un​∥∥vn​∥ summable; every finite quadratic form of XXX is real and nonnegative; its real diagonal is summable with sum 111; and (p0+p1)Re⁡Xpp(p_0+p_1)\operatorname{Re}X_{pp}(p0​+p1​)ReXpp​ is summable. Moreover, there exists m∈R4m\in\mathbb R^4m∈R4 such that for every v∈R4v\in\mathbb R^4v∈R4,

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

each inner family over qqq, and the family of resulting inner sums over ppp, must be 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​.

Finally,

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,​

where Xs,YsX_s,Y_sXs​,Ys​ are coordinate restrictions, including the empty one, g(x)=xlog⁡xg(x)=x\log xg(x)=xlogx, matrix functions use real continuous functional calculus, and [x]+=max⁡(x,0)[x]_+=\max(x,0)[x]+​=max(x,0) is regarded as an extended nonnegative real. There is no assumption that ξ\xiξ is positive: the real logarithm used here has log⁡0=0\log0=0log0=0 and log⁡ξ=log⁡∣ξ∣\log\xi=\log|\xi|logξ=log∣ξ∣ for negative ξ\xiξ; all quotients are total, and real square roots of negative arguments are zero.

QAlgorithms.Papers.Independent.Intakenext100.P69.centeredTwoModeGaussianKernel

For every real matrix VVV indexed by {0,1}×{0,1}\{0,1\}\times\{0,1\}{0,1}×{0,1}, this definition selects an arbitrary complex kernel ρ=(ρij)i,j∈N2\rho=(\rho_{ij})_{i,j\in\mathbb N^2}ρ=(ρij​)i,j∈N2​ satisfying the conditions below if any such kernel exists; it specifies no uniqueness. The conditions are that VT=VV^{\mathsf T}=VVT=V; ρij=∑nun(i)wn(j)‾\rho_{ij}=\sum_nu_n(i)\overline{w_n(j)}ρij​=∑n​un​(i)wn​(j)​ for sequences un,wn∈ℓ2(N2;C)u_n,w_n\in\ell^2(\mathbb N^2;\mathbb C)un​,wn​∈ℓ2(N2;C) with summable (∥un∥∥wn∥)n(\|u_n\|\|w_n\|)_n(∥un​∥∥wn​∥)n​; every finite quadratic form ∑i,j∈sxi‾ρijxj\sum_{i,j\in s}\overline{x_i}\rho_{ij}x_j∑i,j∈s​xi​​ρij​xj​, for finite s⊆N2s\subseteq\mathbb N^2s⊆N2 and arbitrary x:N2→Cx:\mathbb N^2\to\mathbb Cx:N2→C, is real and nonnegative; (Re⁡ρii)i(\operatorname{Re}\rho_{ii})_i(Reρii​)i​ is summable with sum 111; 2V2V2V is positive definite and 2V+iΩ2V+i\Omega2V+iΩ is positive semidefinite, where Ω=diag⁡((01−10),(01−10))\Omega=\operatorname{diag}\bigl(\begin{pmatrix}0&1\\-1&0\end{pmatrix},\begin{pmatrix}0&1\\-1&0\end{pmatrix}\bigr)Ω=diag((0−1​10​),(0−1​10​)) in interleaved mode order; and ((i0+i1)Re⁡ρii)i((i_0+i_1)\operatorname{Re}\rho_{ii})_i((i0​+i1​)Reρii​)i​ is summable. For a real vector v=(vk,a)v=(v_{k,a})v=(vk,a​), define zk=(−vk,1+ivk,0)/2z_k=(-v_{k,1}+iv_{k,0})/\sqrt2zk​=(−vk,1​+ivk,0​)/2​ and

Dz(a,b)=e−∣z∣2/2a! b!∑h=0min⁡(a,b)za−h(−z‾)b−hh! (a−h)! (b−h)!,Tij(v)=ρij∏k=01Dzk(jk,ik).D_z(a,b)=e^{-|z|^2/2}\sqrt{a!\,b!} \sum_{h=0}^{\min(a,b)} \frac{z^{a-h}(-\overline z)^{b-h}}{h!\,(a-h)!\,(b-h)!}, \qquad T_{ij}(v)=\rho_{ij}\prod_{k=0}^{1}D_{z_k}(j_k,i_k).Dz​(a,b)=e−∣z∣2/2a!b!​h=0∑min(a,b)​h!(a−h)!(b−h)!za−h(−z)b−h​,Tij​(v)=ρij​k=0∏1​Dzk​​(jk​,ik​).

There must exist a real four-component vector mmm such that for every real vvv, each inner family (Tij(v))j(T_{ij}(v))_j(Tij​(v))j​ and the outer family (∑jTij(v))i(\sum_jT_{ij}(v))_i(∑j​Tij​(v))i​ are summable and

∑i∑jTij(v)=exp⁡ ⁣(−14vT(2V)v+i m⋅v);\sum_i\sum_jT_{ij}(v) =\exp\!\left(-\frac14v^{\mathsf T}(2V)v+i\,m\cdot v\right);i∑​j∑​Tij​(v)=exp(−41​vT(2V)v+im⋅v);

also, for every vvv, this sum must equal exp⁡(−vT(2V)v/4)\exp(-v^{\mathsf T}(2V)v/4)exp(−vT(2V)v/4). If no kernel satisfies all these conditions, the definition instead returns the kernel whose (i,j)(i,j)(i,j)-entry is 111 when both i=(0,0)i=(0,0)i=(0,0) and j=(0,0)j=(0,0)j=(0,0), and is 000 otherwise.

QAlgorithms.Papers.Independent.Intakenext100.P69.comparisonC

For arbitrary real numbers η,r\eta,rη,r, put

b=−∣η−1∣+ηe2r+e−2r2(η+1−e2r∣η−1∣),a=2b+(η−1)e−2r2η.b=\frac{-|\eta-1|+\eta e^{2r}+e^{-2r}} {2\bigl(\eta+1-e^{2r}|\eta-1|\bigr)},\qquad a=\frac{2b+(\eta-1)e^{-2r}}{2\eta}.b=2(η+1−e2r∣η−1∣)−∣η−1∣+ηe2r+e−2r​,a=2η2b+(η−1)e−2r​.

This definition returns

(a−12)(b−12).\sqrt{\left(a-\frac12\right)\left(b-\frac12\right)}.(a−21​)(b−21​)​.

Division by zero in the formulas for a,ba,ba,b gives 000, and the returned square root is 000 whenever its argument is nonpositive. No hypothesis requires either factor to be nonnegative.

QAlgorithms.Papers.Independent.Intakenext100.P69.comparisonCovariance

For arbitrary real numbers η,r\eta,rη,r, define

b=−∣η−1∣+ηe2r+e−2r2(η+1−e2r∣η−1∣),a=2b+(η−1)e−2r2η,h=(a−12)(b−12).b=\frac{-|\eta-1|+\eta e^{2r}+e^{-2r}} {2\bigl(\eta+1-e^{2r}|\eta-1|\bigr)},\qquad a=\frac{2b+(\eta-1)e^{-2r}}{2\eta},\qquad h=\sqrt{\left(a-\frac12\right)\left(b-\frac12\right)}.b=2(η+1−e2r∣η−1∣)−∣η−1∣+ηe2r+e−2r​,a=2η2b+(η−1)e−2r​,h=(a−21​)(b−21​)​.

This definition returns the real matrix, indexed in the 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),

(a0h00a0−hh0b00−h0b).\begin{pmatrix} a&0&h&0\\ 0&a&0&-h\\ h&0&b&0\\ 0&-h&0&b \end{pmatrix}.​a0h0​0a0−h​h0b0​0−h0b​​.

There are no parameter restrictions; division by zero gives 000, and square roots of negative real numbers give 000. No positive definiteness or quantum covariance condition is asserted.

QAlgorithms.Papers.Independent.Intakenext100.P69.comparisonKernel

For arbitrary real numbers η,r\eta,rη,r, form

b=−∣η−1∣+ηe2r+e−2r2(η+1−e2r∣η−1∣),a=2b+(η−1)e−2r2η,h=(a−12)(b−12),V=(a0h00a0−hh0b00−h0b),b=\frac{-|\eta-1|+\eta e^{2r}+e^{-2r}} {2\bigl(\eta+1-e^{2r}|\eta-1|\bigr)},\qquad a=\frac{2b+(\eta-1)e^{-2r}}{2\eta},\qquad h=\sqrt{\left(a-\frac12\right)\left(b-\frac12\right)},\qquad V=\begin{pmatrix} a&0&h&0\\ 0&a&0&-h\\ h&0&b&0\\ 0&-h&0&b \end{pmatrix},b=2(η+1−e2r∣η−1∣)−∣η−1∣+ηe2r+e−2r​,a=2η2b+(η−1)e−2r​,h=(a−21​)(b−21​)​,V=​a0h0​0a0−h​h0b0​0−h0b​​,

using the 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), totalized division with quotient 000 at a zero denominator, and real square root equal to 000 on negative arguments. The definition selects an arbitrary kernel ρ=(ρij)i,j∈N2\rho=(\rho_{ij})_{i,j\in\mathbb N^2}ρ=(ρij​)i,j∈N2​ satisfying all the following conditions if one exists. The matrix VVV must be symmetric; ρ\rhoρ must admit ρij=∑nun(i)wn(j)‾\rho_{ij}=\sum_nu_n(i)\overline{w_n(j)}ρij​=∑n​un​(i)wn​(j)​ with un,wn∈ℓ2(N2;C)u_n,w_n\in\ell^2(\mathbb N^2;\mathbb C)un​,wn​∈ℓ2(N2;C) and summable (∥un∥∥wn∥)n(\|u_n\|\|w_n\|)_n(∥un​∥∥wn​∥)n​; every finite quadratic form ∑i,j∈sxi‾ρijxj\sum_{i,j\in s}\overline{x_i}\rho_{ij}x_j∑i,j∈s​xi​​ρij​xj​, for finite s⊆N2s\subseteq\mathbb N^2s⊆N2 and arbitrary x:N2→Cx:\mathbb N^2\to\mathbb Cx:N2→C, must be real and nonnegative; the real diagonal must be summable with sum 111; 2V2V2V must be positive definite and 2V+iΩ2V+i\Omega2V+iΩ positive semidefinite, where Ω=diag⁡(J,J)\Omega=\operatorname{diag}(J,J)Ω=diag(J,J) and J=(01−10)J=\begin{pmatrix}0&1\\-1&0\end{pmatrix}J=(0−1​10​); and ((i0+i1)Re⁡ρii)i((i_0+i_1)\operatorname{Re}\rho_{ii})_i((i0​+i1​)Reρii​)i​ must be summable. For each real four-component vector vvv, set zk=(−vk,1+ivk,0)/2z_k=(-v_{k,1}+iv_{k,0})/\sqrt2zk​=(−vk,1​+ivk,0​)/2​ and

Dz(p,q)=e−∣z∣2/2p! q!∑ℓ=0min⁡(p,q)zp−ℓ(−z‾)q−ℓℓ! (p−ℓ)! (q−ℓ)!,Tij(v)=ρij∏k=01Dzk(jk,ik).D_z(p,q)=e^{-|z|^2/2}\sqrt{p!\,q!} \sum_{\ell=0}^{\min(p,q)} \frac{z^{p-\ell}(-\overline z)^{q-\ell}}{\ell!\,(p-\ell)!\,(q-\ell)!}, \qquad T_{ij}(v)=\rho_{ij}\prod_{k=0}^{1}D_{z_k}(j_k,i_k).Dz​(p,q)=e−∣z∣2/2p!q!​ℓ=0∑min(p,q)​ℓ!(p−ℓ)!(q−ℓ)!zp−ℓ(−z)q−ℓ​,Tij​(v)=ρij​k=0∏1​Dzk​​(jk​,ik​).

There must exist a real four-component vector mmm such that for every vvv, each family (Tij(v))j(T_{ij}(v))_j(Tij​(v))j​ and the family (∑jTij(v))i(\sum_jT_{ij}(v))_i(∑j​Tij​(v))i​ are summable and

∑i∑jTij(v)=exp⁡ ⁣(−14vT(2V)v+i m⋅v);\sum_i\sum_jT_{ij}(v) =\exp\!\left(-\frac14v^{\mathsf T}(2V)v+i\,m\cdot v\right);i∑​j∑​Tij​(v)=exp(−41​vT(2V)v+im⋅v);

additionally, for every vvv, this sum must equal exp⁡(−vT(2V)v/4)\exp(-v^{\mathsf T}(2V)v/4)exp(−vT(2V)v/4). No uniqueness is specified. If no kernel satisfies these conditions, the returned kernel is 111 at i=j=(0,0)i=j=(0,0)i=j=(0,0) and 000 at every other pair of indices.

QAlgorithms.Papers.Independent.Intakenext100.P69.gaussianComparisonBound

For arbitrary real η,r\eta,rη,r, this defines

G(η,r)=Dbits(K(Vη,r),K(Wη,r)),G(\eta,r)=D_{\mathrm{bits}}\bigl(K(V_{\eta,r}),K(W_{\eta,r})\bigr),G(η,r)=Dbits​(K(Vη,r​),K(Wη,r​)),

using pair indices Q=N2Q=\mathbb N^2Q=N2, which identify a pair with the function on {0,1}\{0,1\}{0,1} having those two values. Set

d=η+1−e2r∣η−1∣,b=−∣η−1∣+ηe2r+e−2r2d,a=2b+(η−1)e−2r2η,c=2b−e−2r2η,c∗=(a−12)(b−12),Vη,r=(a0c00a0−cc0b00−c0b),Wη,r=(a0c∗00a0−c∗c∗0b00−c∗0b).\begin{gathered} d=\eta+1-e^{2r}|\eta-1|, \qquad b=\frac{-|\eta-1|+\eta e^{2r}+e^{-2r}}{2d}, \qquad a=\frac{2b+(\eta-1)e^{-2r}}{2\eta},\\ c=\frac{2b-e^{-2r}}{2\sqrt\eta}, \qquad c_*=\sqrt{(a-\tfrac12)(b-\tfrac12)},\\ V_{\eta,r}= \begin{pmatrix} a&0&c&0\\ 0&a&0&-c\\ c&0&b&0\\ 0&-c&0&b \end{pmatrix}, \qquad W_{\eta,r}= \begin{pmatrix} a&0&c_*&0\\ 0&a&0&-c_*\\ c_*&0&b&0\\ 0&-c_*&0&b \end{pmatrix}. \end{gathered}d=η+1−e2r∣η−1∣,b=2d−∣η−1∣+ηe2r+e−2r​,a=2η2b+(η−1)e−2r​,c=2η​2b−e−2r​,c∗​=(a−21​)(b−21​)​,Vη,r​=​a0c0​0a0−c​c0b0​0−c0b​​,Wη,r​=​a0c∗​0​0a0−c∗​​c∗​0b0​0−c∗​0b​​.​

The coordinate order is (0,0),(0,1),(1,0),(1,1)(0,0),(0,1),(1,0),(1,1)(0,0),(0,1),(1,0),(1,1). For a real 4×44\times44×4 matrix UUU, K(U)K(U)K(U) is the definition’s chosen kernel XXX satisfying all of the following conditions, if such a kernel exists, and otherwise is the vacuum kernel Xpq=1X_{pq}=1Xpq​=1 when p=q=(0,0)p=q=(0,0)p=q=(0,0) and 000 otherwise: UT=UU^{\mathsf T}=UUT=U; XXX admits an expansion Xpq=∑nun(p)vn(q)‾X_{pq}=\sum_n u_n(p)\overline{v_n(q)}Xpq​=∑n​un​(p)vn​(q)​ with un,vn∈ℓ2(Q;C)u_n,v_n\in\ell^2(Q;\mathbb C)un​,vn​∈ℓ2(Q;C) and ∑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; ∑pRe⁡Xpp\sum_p\operatorname{Re}X_{pp}∑p​ReXpp​ is summable and equals 111; the family (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) and J=(01−10)J=\left(\begin{smallmatrix}0&1\\-1&0\end{smallmatrix}\right)J=(0−1​10​). In addition, 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∈Q∑q∈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\in Q}\sum_{q\in 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∑​q∈Q∑​Xpq​j=0∏1​Dzj​​(qj​,pj​),zj​=2​−vj,1​+ivj,0​​.

For each v,pv,pv,p, the inner summand family over qqq must be summable, and the resulting family of inner sums over ppp must also be summable. The displacement entries used here are

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​.

The relative-entropy quantity in the definition is

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

where coordinate restrictions include the empty restriction, g(x)=xlog⁡xg(x)=x\log xg(x)=xlogx, matrix functions use real continuous functional calculus, and [x]+=max⁡(x,0)[x]_+=\max(x,0)[x]+​=max(x,0) is embedded in [0,+∞][0,+\infty][0,+∞]. No parameter restrictions are imposed: zero denominators give zero quotients, and the real square root of a negative argument is zero.

QAlgorithms.Papers.Independent.Intakenext100.P69.normalStateRelativeEntropyBits

For every type III, with no finiteness, countability, or nonemptiness assumption, and every two complex matrices ρ,σ\rho,\sigmaρ,σ indexed by III, this defines an extended nonnegative real number

Dbits(ρ,σ)=1log⁡2sup⁡s⊆Is finited(ρs,σs),d(X,Y)={[Re⁡Tr⁡(g(X)−Xlog⁡Y+Y−X)]+,ker⁡Y⊆ker⁡X,+∞,otherwise.D_{\mathrm{bits}}(\rho,\sigma) =\frac{1}{\log 2}\sup_{\substack{s\subseteq I\\s\text{ finite}}}d(\rho_s,\sigma_s), \qquad d(X,Y)= \begin{cases} \bigl[\operatorname{Re}\operatorname{Tr}\bigl(g(X)-X\log Y+Y-X\bigr)\bigr]_+, &\ker Y\subseteq\ker X,\\ +\infty,&\text{otherwise}. \end{cases}Dbits​(ρ,σ)=log21​s⊆Is finite​sup​d(ρs​,σs​),d(X,Y)={[ReTr(g(X)−XlogY+Y−X)]+​,+∞,​kerY⊆kerX,otherwise.​

Here ρs,σs\rho_s,\sigma_sρs​,σs​ are the matrices restricted to rows and columns in sss; the kernels are those of their finite-dimensional matrix multiplication maps; g(x)=xlog⁡xg(x)=x\log xg(x)=xlogx; g(X)g(X)g(X) and log⁡Y\log YlogY use real continuous functional calculus; and [a]+=max⁡(a,0)[a]_+=\max(a,0)[a]+​=max(a,0), viewed as an extended nonnegative real. The supremum includes s=∅s=\varnothings=∅, and the division is in the extended nonnegative reals by the positive finite number log⁡2\log 2log2. Neither matrix is assumed to be a density matrix, positive, or trace class; the definition also applies when III is empty.

QAlgorithms.Papers.Independent.Intakenext100.P69.relativeEntropyEntanglementBits

For every complex matrix ρ\rhoρ indexed by Q=N2Q=\mathbb N^2Q=N2, without any assumption on ρ\rhoρ, this defines

Ebits(ρ)=inf⁡σDbits(ρ,σ),E_{\mathrm{bits}}(\rho) =\inf_{\sigma}D_{\mathrm{bits}}(\rho,\sigma),Ebits​(ρ)=σinf​Dbits​(ρ,σ),

where the infimum ranges over precisely the following matrices σ\sigmaσ. First, σ\sigmaσ is a density kernel: it has a nuclear expansion σij=∑n=0∞un(i)vn(j)‾\sigma_{ij}=\sum_{n=0}^{\infty}u_n(i)\overline{v_n(j)}σij​=∑n=0∞​un​(i)vn​(j)​ with un,vn∈ℓ2(Q;C)u_n,v_n\in\ell^2(Q;\mathbb C)un​,vn​∈ℓ2(Q;C) and ∑n∥un∥∥vn∥\sum_n\|u_n\|\|v_n\|∑n​∥un​∥∥vn​∥ summable; for every finite s⊆Qs\subseteq Qs⊆Q and every z:Q→Cz:Q\to\mathbb Cz:Q→C, the number ∑i,j∈szi‾σijzj\sum_{i,j\in s}\overline{z_i}\sigma_{ij}z_j∑i,j∈s​zi​​σij​zj​ is real and nonnegative; and the real diagonal entries form a summable family with sum 111. Second, for every real ε>0\varepsilon>0ε>0, there is a finite convex combination YYY of product kernels R⊗SR\otimes SR⊗S, where R,SR,SR,S are density kernels on N\mathbb NN satisfying these same conditions and

(R⊗S)(i,a),(j,b)=RijSab,τ(σ−Y)<ε.(R\otimes S)_{(i,a),(j,b)}=R_{ij}S_{ab}, \qquad \tau(\sigma-Y)<\varepsilon.(R⊗S)(i,a),(j,b)​=Rij​Sab​,τ(σ−Y)<ε.

Here

τ(X)=inf⁡Xij=∑nun(i)vn(j)‾un,vn∈ℓ2(Q;C),∑n∥un∥∥vn∥ summable[∑n∥un∥∥vn∥]+,\tau(X)= \inf_{\substack{X_{ij}=\sum_n u_n(i)\overline{v_n(j)}\\ u_n,v_n\in\ell^2(Q;\mathbb C),\;\sum_n\|u_n\|\|v_n\|\text{ summable}}} \left[\sum_n\|u_n\|\|v_n\|\right]_+,τ(X)=Xij​=∑n​un​(i)vn​(j)​un​,vn​∈ℓ2(Q;C),∑n​∥un​∥∥vn​∥ summable​inf​[n∑​∥un​∥∥vn​∥]+​,

with an empty infimum equal to +∞+\infty+∞. Finally, the quantity minimized is

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

where Xs,YsX_s,Y_sXs​,Ys​ are coordinate restrictions, g(x)=xlog⁡xg(x)=x\log xg(x)=xlogx, the matrix functions use real continuous functional calculus, and [a]+=max⁡(a,0)[a]_+=\max(a,0)[a]+​=max(a,0) is embedded in the extended nonnegative reals. All these infima and suprema are taken in [0,+∞][0,+\infty][0,+∞]; the finite-coordinate supremum includes the empty set.

QAlgorithms.Papers.Independent.Intakenext100.P69.resourceA

For arbitrary real numbers η\etaη and rrr, put

b=−∣η−1∣+ηe2r+e−2r2(η+1−e2r∣η−1∣).b=\frac{-|\eta-1|+\eta e^{2r}+e^{-2r}} {2\bigl(\eta+1-e^{2r}|\eta-1|\bigr)}.b=2(η+1−e2r∣η−1∣)−∣η−1∣+ηe2r+e−2r​.

This definition returns

2b+(η−1)e−2r2η.\frac{2b+(\eta-1)e^{-2r}}{2\eta}.2η2b+(η−1)e−2r​.

Both divisions are totalized: a zero denominator gives quotient 000. In particular, the returned value is 000 when η=0\eta=0η=0; no positivity or denominator hypothesis is imposed.

QAlgorithms.Papers.Independent.Intakenext100.P69.resourceB

For arbitrary real numbers η\etaη and rrr, this defines

−∣η−1∣+ηe2r+e−2r2(η+1−e2r∣η−1∣).\frac{-|\eta-1|+\eta e^{2r}+e^{-2r}} {2\bigl(\eta+1-e^{2r}|\eta-1|\bigr)}.2(η+1−e2r∣η−1∣)−∣η−1∣+ηe2r+e−2r​.

No condition excludes a zero denominator; real division is totalized, so the defined value is 000 whenever that denominator is zero.

QAlgorithms.Papers.Independent.Intakenext100.P69.resourceC

For arbitrary real numbers η\etaη and rrr, put

b=−∣η−1∣+ηe2r+e−2r2(η+1−e2r∣η−1∣).b=\frac{-|\eta-1|+\eta e^{2r}+e^{-2r}} {2\bigl(\eta+1-e^{2r}|\eta-1|\bigr)}.b=2(η+1−e2r∣η−1∣)−∣η−1∣+ηe2r+e−2r​.

This definition returns

2b−e−2r2η.\frac{2b-e^{-2r}}{2\sqrt{\eta}}.2η​2b−e−2r​.

Here the real square root is 000 on negative arguments, and division by zero gives 000; these conventions also apply to the definition of bbb. Consequently, the returned value is 000 for every η≤0\eta\le0η≤0, with no restriction on rrr.

QAlgorithms.Papers.Independent.Intakenext100.P69.resourceCovariance

For arbitrary real numbers η,r\eta,rη,r, define

b=−∣η−1∣+ηe2r+e−2r2(η+1−e2r∣η−1∣),a=2b+(η−1)e−2r2η,c=2b−e−2r2η.b=\frac{-|\eta-1|+\eta e^{2r}+e^{-2r}} {2\bigl(\eta+1-e^{2r}|\eta-1|\bigr)},\qquad a=\frac{2b+(\eta-1)e^{-2r}}{2\eta},\qquad c=\frac{2b-e^{-2r}}{2\sqrt{\eta}}.b=2(η+1−e2r∣η−1∣)−∣η−1∣+ηe2r+e−2r​,a=2η2b+(η−1)e−2r​,c=2η​2b−e−2r​.

This definition returns the real matrix, indexed in the 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),

(a0c00a0−cc0b00−c0b).\begin{pmatrix} a&0&c&0\\ 0&a&0&-c\\ c&0&b&0\\ 0&-c&0&b \end{pmatrix}.​a0c0​0a0−c​c0b0​0−c0b​​.

All parameters are allowed, division by zero gives 000, and the real square root of a negative number is 000. The definition itself imposes no positive definiteness or quantum covariance condition.

QAlgorithms.Papers.Independent.Intakenext100.P69.resourceDenominator

For arbitrary real numbers η\etaη and rrr, this defines the real number

η+1−e2r∣η−1∣.\eta+1-e^{2r}|\eta-1|.η+1−e2r∣η−1∣.

There are no restrictions on either parameter, and this value may be positive, zero, or negative.

QAlgorithms.Papers.Independent.Intakenext100.P69.resourceKernel

For arbitrary real numbers η,r\eta,rη,r, form

b=−∣η−1∣+ηe2r+e−2r2(η+1−e2r∣η−1∣),a=2b+(η−1)e−2r2η,c=2b−e−2r2η,V=(a0c00a0−cc0b00−c0b),b=\frac{-|\eta-1|+\eta e^{2r}+e^{-2r}} {2\bigl(\eta+1-e^{2r}|\eta-1|\bigr)},\qquad a=\frac{2b+(\eta-1)e^{-2r}}{2\eta},\qquad c=\frac{2b-e^{-2r}}{2\sqrt{\eta}},\qquad V=\begin{pmatrix} a&0&c&0\\ 0&a&0&-c\\ c&0&b&0\\ 0&-c&0&b \end{pmatrix},b=2(η+1−e2r∣η−1∣)−∣η−1∣+ηe2r+e−2r​,a=2η2b+(η−1)e−2r​,c=2η​2b−e−2r​,V=​a0c0​0a0−c​c0b0​0−c0b​​,

using the 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), totalized division with quotient 000 at a zero denominator, and real square root equal to 000 on negative arguments. The definition selects an arbitrary kernel ρ=(ρij)i,j∈N2\rho=(\rho_{ij})_{i,j\in\mathbb N^2}ρ=(ρij​)i,j∈N2​ satisfying all the following conditions if one exists. The matrix VVV must be symmetric; ρ\rhoρ must admit ρij=∑nun(i)wn(j)‾\rho_{ij}=\sum_nu_n(i)\overline{w_n(j)}ρij​=∑n​un​(i)wn​(j)​ with un,wn∈ℓ2(N2;C)u_n,w_n\in\ell^2(\mathbb N^2;\mathbb C)un​,wn​∈ℓ2(N2;C) and summable (∥un∥∥wn∥)n(\|u_n\|\|w_n\|)_n(∥un​∥∥wn​∥)n​; every finite quadratic form ∑i,j∈sxi‾ρijxj\sum_{i,j\in s}\overline{x_i}\rho_{ij}x_j∑i,j∈s​xi​​ρij​xj​, for finite s⊆N2s\subseteq\mathbb N^2s⊆N2 and arbitrary x:N2→Cx:\mathbb N^2\to\mathbb Cx:N2→C, must be real and nonnegative; the real diagonal must be summable with sum 111; 2V2V2V must be positive definite and 2V+iΩ2V+i\Omega2V+iΩ positive semidefinite, where Ω=diag⁡(J,J)\Omega=\operatorname{diag}(J,J)Ω=diag(J,J) and J=(01−10)J=\begin{pmatrix}0&1\\-1&0\end{pmatrix}J=(0−1​10​); and ((i0+i1)Re⁡ρii)i((i_0+i_1)\operatorname{Re}\rho_{ii})_i((i0​+i1​)Reρii​)i​ must be summable. For each real four-component vector vvv, set zk=(−vk,1+ivk,0)/2z_k=(-v_{k,1}+iv_{k,0})/\sqrt2zk​=(−vk,1​+ivk,0​)/2​ and

Dz(p,q)=e−∣z∣2/2p! q!∑h=0min⁡(p,q)zp−h(−z‾)q−hh! (p−h)! (q−h)!,Tij(v)=ρij∏k=01Dzk(jk,ik).D_z(p,q)=e^{-|z|^2/2}\sqrt{p!\,q!} \sum_{h=0}^{\min(p,q)} \frac{z^{p-h}(-\overline z)^{q-h}}{h!\,(p-h)!\,(q-h)!}, \qquad T_{ij}(v)=\rho_{ij}\prod_{k=0}^{1}D_{z_k}(j_k,i_k).Dz​(p,q)=e−∣z∣2/2p!q!​h=0∑min(p,q)​h!(p−h)!(q−h)!zp−h(−z)q−h​,Tij​(v)=ρij​k=0∏1​Dzk​​(jk​,ik​).

There must exist a real four-component vector mmm such that for every vvv, each family (Tij(v))j(T_{ij}(v))_j(Tij​(v))j​ and the family (∑jTij(v))i(\sum_jT_{ij}(v))_i(∑j​Tij​(v))i​ are summable and

∑i∑jTij(v)=exp⁡ ⁣(−14vT(2V)v+i m⋅v);\sum_i\sum_jT_{ij}(v) =\exp\!\left(-\frac14v^{\mathsf T}(2V)v+i\,m\cdot v\right);i∑​j∑​Tij​(v)=exp(−41​vT(2V)v+im⋅v);

additionally, for every vvv, this sum must equal exp⁡(−vT(2V)v/4)\exp(-v^{\mathsf T}(2V)v/4)exp(−vT(2V)v/4). No uniqueness is specified. If no kernel satisfies these conditions, the returned kernel is 111 at i=j=(0,0)i=j=(0,0)i=j=(0,0) and 000 at every other pair of indices.

QAlgorithms.Papers.Independent.Intakenext100.P69.resourceNoise

For arbitrary real numbers η\etaη and rrr, this defines the real number

e−2r(η+1)2.\frac{e^{-2r}(\eta+1)}{2}.2e−2r(η+1)​.

No nonnegativity condition is imposed: the value is zero for η=−1\eta=-1η=−1 and negative for η<−1\eta<-1η<−1.

QAlgorithms.Papers.Independent.Intakenext100.P69.twoModeSplit

For every complex matrix ρ\rhoρ whose rows and columns are indexed by functions {0,1}→N\{0,1\}\to\mathbb N{0,1}→N, this defines a complex matrix indexed by N2\mathbb N^2N2 by relabeling those functions as ordered pairs:

twoModeSplit⁡(ρ)(p0,p1),(q0,q1)=ρ i↦pi,  i↦qi.\operatorname{twoModeSplit}(\rho)_{(p_0,p_1),(q_0,q_1)} =\rho_{\,i\mapsto p_i,\;i\mapsto q_i}.twoModeSplit(ρ)(p0​,p1​),(q0​,q1​)​=ρi↦pi​,i↦qi​​.

Thus the row function takes the value p0p_0p0​ at 000 and p1p_1p1​ at 111, and the column function similarly takes the values q0,q1q_0,q_1q0​,q1​. There is no positivity, normalization, summability, or Gaussianity assumption on ρ\rhoρ.

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