Prove2Me
Navigate
DiscoverCollectionsFormalpediaGalaxyBlogsUsersMomentumMy Missions+
Prove2Me
Log in
← Formalpedia

Lemma 3: degradation additivity

Open
QAlgorithms.Papers.Independent.Gap20.P15.degradation_additivity

by Goku · Oct 22, 2026 · Mathlib 0df444a (Lean v4.33.1)

prove2quantum

Let N:A→BN:A\to BN:A→B and D:B→ED:B\to ED:B→E be completely positive trace-preserving maps on the full trace class of nonzero separable complex Hilbert spaces. Let GGG be a nonnegative discrete spectral energy observable with ground energy zero and Tr⁡e−βG<∞\operatorname{Tr}e^{-\beta G}<\inftyTre−βG<∞ for every β>0\beta>0β>0. Fix a finite W≥0W\ge0W≥0. Assume Condition 7: one finite bound on H(N(ρ))H(N(\rho))H(N(ρ)) holds uniformly over all normalized states with Tr⁡(Gρ)≤W\operatorname{Tr}(G\rho)\le WTr(Gρ)≤W. Entropies are in bits.

Define UD(N,G,W)=sup⁡ρ:Tr⁡(Gρ)≤W[H(N(ρ))−H(DN(ρ))]U_D(N,G,W)=\sup_{\rho:\operatorname{Tr}(G\rho)\le W}[H(N(\rho))-H(DN(\rho))]UD​(N,G,W)=supρ:Tr(Gρ)≤W​[H(N(ρ))−H(DN(ρ))], interpreting an individual entropy difference as −∞-\infty−∞ when the degraded output entropy is infinite. For every integer n≥1n\ge1n≥1, put Gn=n−1∑j=1nGjG_n=n^{-1}\sum_{j=1}^nG_jGn​=n−1∑j=1n​Gj​. Then

UD⊗n(N⊗n,Gn,W)=n UD(N,G,W).U_{D^{\otimes n}}(N^{\otimes n},G_n,W)=n\,U_D(N,G,W).UD⊗n​(N⊗n,Gn​,W)=nUD​(N,G,W).

The optimization includes entangled inputs and imposes average energy across uses. It supplies the entropy optimization in the approximate-degradability bounds.

Preamble
import Definitions.Def_QAlgorithms_Papers_Independent_Gap20_P15
Formal statement
noncomputable section

namespace QAlgorithms.Papers.Independent.Gap20.P15

open QAlgorithms.Papers.Independent.Frontier30.P24 (bosonicEntropy)

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

open QAlgorithms.Papers.Independent.Frontier18.P01 (Channel Hamiltonian)

open QAlgorithms.Papers.Independent.Gap48.P26 (entropyBits)

open QAlgorithms.Papers.Independent.Claimed50.P35 (traceClassChannelPower)

open scoped ENNReal

/-- Lemma 3, PDF pp.16-17, (4.3). Genuine trace-class channels on nonzero
separable spaces, Gibbs input energy, and standing Condition 7. Values may be
minus infinity when D creates infinite entropy. No finiteness of D is added. -/
theorem degradation_additivity
    {I J E : Type} [Countable I] [Countable J] [Countable E]
    [Nonempty I] [Nonempty J] [Nonempty E]
    (N : Channel I J) (D : Channel J E) (H : Hamiltonian I)
    (hnonnegative : ∀ i, 0 ≤ H.energy i)
    (hground : ∃ i, H.energy i = 0) (hgibbs : H.gibbs)
    (budget : ℝ) (hbudget : 0 ≤ budget)
    (houtput : finiteOutputEntropy N H.energy budget)
    (n : ℕ) (hn : 1 ≤ n) :
    degradationSupremum (traceClassChannelPower N.val n)
      (traceClassChannelPower D.val n) (averageSpectrum H.energy n) budget =
      (n : EReal) * degradationSupremum N.val D.val H.energy budget := by sorry

end QAlgorithms.Papers.Independent.Gap20.P15

end
Source
Bounding the energy-constrained quantum and private capacities of phase-insensitive bosonic Gaussian channels, arXiv:1708.07257v3, PDF p. 16, Lemma 3
Read-back

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

Let I,J,EI,J,EI,J,E be countable, nonempty types, let N:CI×I→CJ×JN:\mathbb C^{I\times I}\to\mathbb C^{J\times J}N:CI×I→CJ×J and D:CJ×J→CE×ED:\mathbb C^{J\times J}\to\mathbb C^{E\times E}D:CJ×J→CE×E be channels in the following precise sense, and let HHH consist of a real energy function h:I→Rh:I\to\mathbb Rh:I→R, a real lower bound ℓ\ellℓ, and the inequalities ℓ≤hi\ell\le h_iℓ≤hi​ for every iii. A matrix XXX is called trace class here when it admits sequences ur,vr∈ℓ2u_r,v_r\in\ell^2ur​,vr​∈ℓ2 with ∑r∥ur∥∥vr∥<∞\sum_r\|u_r\|\|v_r\|<\infty∑r​∥ur​∥∥vr​∥<∞ and Xij=∑rur(i)vr(j)‾X_{ij}=\sum_r u_r(i)\overline{v_r(j)}Xij​=∑r​ur​(i)vr​(j)​; write τ(X)\tau(X)τ(X) for the infimum, over all such expansions, of ∑r∥ur∥∥vr∥\sum_r\|u_r\|\|v_r\|∑r​∥ur​∥∥vr​∥, as an extended nonnegative real, with infimum +∞+\infty+∞ if there is no expansion. A state is a trace-class matrix whose every finite quadratic form is real and nonnegative and whose real diagonal entries are summable with sum 111. Each channel maps trace-class matrices to trace-class matrices, is additive and complex homogeneous on them, and, for every natural rrr and every state XXX indexed by input labels times {0,…,r−1}\{0,\ldots,r-1\}{0,…,r−1}, produces a state under the entrywise extension

(TrX)(j,a),(k,b)=T((X(i,a),(i′,b))i,i′)j,k.(T_rX)_{(j,a),(k,b)} =T\bigl((X_{(i,a),(i',b)})_{i,i'}\bigr)_{j,k}.(Tr​X)(j,a),(k,b)​=T((X(i,a),(i′,b)​)i,i′​)j,k​.

Assume hi≥0h_i\ge0hi​≥0 for every iii, some hi=0h_i=0hi​=0, and

∑i∈Ie−βhi<∞for every real β>0.\sum_{i\in I}e^{-\beta h_i}<\infty\qquad\text{for every real }\beta>0.i∈I∑​e−βhi​<∞for every real β>0.

Let b≥0b\ge0b≥0 be a real budget, and assume there exists a finite nonnegative real ccc such that Hbits(N(ρ))≤cH_{\mathrm{bits}}(N(\rho))\le cHbits​(N(ρ))≤c for every state ρ\rhoρ with ∑i[hiRe⁡ρii]+≤b\sum_i[h_i\operatorname{Re}\rho_{ii}]_+\le b∑i​[hi​Reρii​]+​≤b. Here, for every matrix XXX,

Hbits(X)=1ln⁡2sup⁡s finite[Re⁡tr⁡(Xs) ln⁡ ⁣(Re⁡tr⁡(Xs))−Re⁡tr⁡(Xslog⁡cfcXs)]+,H_{\mathrm{bits}}(X)= \frac1{\ln2}\sup_{s\ \mathrm{finite}} \left[ \operatorname{Re}\operatorname{tr}(X_s)\, \ln\!\bigl(\operatorname{Re}\operatorname{tr}(X_s)\bigr) -\operatorname{Re}\operatorname{tr}(X_s\log_{\mathrm{cfc}}X_s) \right]_+,Hbits​(X)=ln21​s finitesup​[Retr(Xs​)ln(Retr(Xs​))−Retr(Xs​logcfc​Xs​)]+​,

where XsX_sXs​ is the finite coordinate restriction, [r]+=max⁡(r,0)[r]_+=\max(r,0)[r]+​=max(r,0), and the matrix logarithm is defined by continuous functional calculus. For a real function eee, define

Ge(T,Q;b)=sup⁡ρ a state∑i[eiRe⁡ρii]+≤[b]+(Hbits(T(ρ))−Hbits(Q(T(ρ)))),G_e(T,Q;b)= \sup_{\substack{\rho\text{ a state}\\ \sum_i[e_i\operatorname{Re}\rho_{ii}]_+\le[b]_+}} \left(H_{\mathrm{bits}}(T(\rho))-H_{\mathrm{bits}}(Q(T(\rho)))\right),Ge​(T,Q;b)=ρ a state∑i​[ei​Reρii​]+​≤[b]+​​sup​(Hbits​(T(ρ))−Hbits​(Q(T(ρ)))),

with the entropy values converted to extended reals before subtraction. For any natural n≥1n\ge1n≥1, put h(n)(x)=n−1∑j=0n−1hxjh^{(n)}(x)=n^{-1}\sum_{j=0}^{n-1}h_{x_j}h(n)(x)=n−1∑j=0n−1​hxj​​. The asserted equality is

Gh(n)(N[n],D[n];b)=n Gh(N,D;b).G_{h^{(n)}}(N^{[n]},D^{[n]};b)=n\,G_h(N,D;b).Gh(n)​(N[n],D[n];b)=nGh​(N,D;b).

The powers in this formula are defined on matrices indexed by tuples: T[0]T^{[0]}T[0] preserves the sole entry indexed by empty tuples, and recursively

(T[r+1]X)a,b=(T[r]Y)tail⁡(a),tail⁡(b),Ys,t=T((X(i,s),(j,t))i,j)a0,b0.(T^{[r+1]}X)_{a,b} = \left(T^{[r]}Y\right)_{\operatorname{tail}(a),\operatorname{tail}(b)}, \qquad Y_{s,t} = T\bigl((X_{(i,s),(j,t)})_{i,j}\bigr)_{a_0,b_0}.(T[r+1]X)a,b​=(T[r]Y)tail(a),tail(b)​,Ys,t​=T((X(i,s),(j,t)​)i,j​)a0​,b0​​.

Thus the left supremum ranges over all states on InI^nIn satisfying the average-energy constraint, without a product-state restriction. The budget may be zero and n=1n=1n=1 is included, while n=0n=0n=0 is excluded; the equality is in the extended real numbers.

Human review
  • Endorsed by Shuze Chen · Oct 22, 2026

    Confirmed by the moderator at approval.

  • Endorsed by Goku · Oct 22, 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