Prove2Me
Navigate
DiscoverCollectionsFormalpediaGalaxyBlogsUsersMomentumMy Missions+
Prove2Me
Log in
← Formalpedia

Proposition 4: strict Holevo loss under partial trace, discrete and continuous cases

Open
QAlgorithms.Papers.Independent.Gap48.P45.partialTrace_strictHolevoLoss

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

prove2quantum

All physical Hilbert spaces are nonzero complex separable Hilbert spaces, with finite dimensions included. States are positive trace-class operators of trace one, channels are completely positive and trace preserving on the entire trace class, and logarithms are natural. Inner products are conjugate-linear in the first argument. Let HA=HB⊗HEH_A=H_B\otimes H_EHA​=HB​⊗HE​ and Φ=Tr⁡HE\Phi=\operatorname{Tr}_{H_E}Φ=TrHE​​. For a discrete ensemble e={πi,ρi}e=\{\pi_i,\rho_i\}e={πi​,ρi​} with faithful average and finite χ(e)\chi(e)χ(e), assume a single finite integer r<dim⁡HEr<\dim H_Er<dimHE​ bounds every constituent rank. For a generalized Borel ensemble μ\muμ with faithful average and finite χ(μ)\chi(\mu)χ(μ), assume a single such finite rrr bounds ranks on the entire topological support. Respectively,

χ(Φ(e))<χ(e),χ(Φ(μ))<χ(μ).\chi(\Phi(e))<\chi(e),\qquad\chi(\Phi(\mu))<\chi(\mu).χ(Φ(e))<χ(e),χ(Φ(μ))<χ(μ).

These are two separate cases. The dimension is that of the discarded factor; when it is one, the rank hypotheses are impossible for normalized states.

This result gives strict information loss under a partial trace at a faithful ensemble average.

Preamble
import Definitions.Def_QAlgorithms_Papers_Independent_Gap48_P45
Formal statement
noncomputable section

open Filter MeasureTheory

open scoped BigOperators ENNReal Topology

attribute [local instance] Classical.propDecidable Classical.decEq

attribute [local instance] IsStarNormal.instContinuousFunctionalCalculus

namespace QAlgorithms.Papers.Independent.Gap48.P45

open QAlgorithms.RealRegisters (Hilbert)

open QAlgorithms.Papers.Independent.Dependency50.P33
  (State IsTraceClass IsTraceClassChannel traceSize traceClosure rankOne productCoeff tensorApply)

open QAlgorithms.Papers.Independent.Frontier18.P01 (Channel Complementary traceFirst traceSecond)

open QAlgorithms.Papers.Independent.Frontier18.P07
  (GeneralizedEnsemble averageCoefficients extendedEntropyNats extendedRelativeEntropyNats ensembleHolevo pureStates finiteEnsembleCapacity)

open QAlgorithms.Papers.Independent.Frontier18.P18
  (operatorCoeff boundedOperator conjugateCoeff ReversibleOn IsometricallyEquivalent
   IsParsevalFrame IsKrausRepresentation)

open QAlgorithms.Papers.Independent.Claimed50.P16 (tensorRankLE)

open QAlgorithms.Papers.Independent.Next46.P31 (positiveSquareRoot)

open QAlgorithms.Papers.Independent.Next46.P42 (minimumOutputEntropy)

/-- Proposition 4, PDF p.22, BOTH A and B. The bound is uniformly finite and
uses the discarded E dimension; B imposes it on topological support. -/
theorem partialTrace_strictHolevoLoss
    {B E : Type} [Countable B] [Countable E] [Nonempty B] [Nonempty E] :
    (∀ e : DiscreteEnsemble (B × E), FullRank (discreteAverageCoefficients e) →
      (∃ r : ℕ, (r : ℕ∞) < dimensionCount E ∧
        ∀ i, operatorRank (boundedOperator (e.state i).coefficients) ≤ r) →
      discreteHolevo (id : Matrix (B × E) (B × E) ℂ → Matrix (B × E) (B × E) ℂ) e < ⊤ →
      discreteHolevo (traceSecond (I := B) (J := E)) e <
        discreteHolevo (id : Matrix (B × E) (B × E) ℂ → Matrix (B × E) (B × E) ℂ) e) ∧
    (∀ mu : GeneralizedEnsemble (B × E), FullRank (averageCoefficients mu) →
      (∃ r : ℕ, (r : ℕ∞) < dimensionCount E ∧
        ∀ rho ∈ (mu : Measure (State (B × E))).support,
          operatorRank (boundedOperator rho.coefficients) ≤ r) →
      ensembleHolevo (id : Matrix (B × E) (B × E) ℂ → Matrix (B × E) (B × E) ℂ) mu < ⊤ →
      ensembleHolevo (traceSecond (I := B) (J := E)) mu <
        ensembleHolevo (id : Matrix (B × E) (B × E) ℂ → Matrix (B × E) (B × E) ℂ) mu) := by sorry

end QAlgorithms.Papers.Independent.Gap48.P45

end
Source
Reversibility conditions for quantum channels and their applications, arXiv:1203.0262v3, PDF p. 22, Proposition 4: strict Holevo loss under partial trace, discrete and continuous cases
Read-back

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

Let B,EB,EB,E be countable, nonempty types and HA=ℓ2(A;C)H_A=\ell^2(A;\mathbb C)HA​=ℓ2(A;C). A trace-class coefficient matrix admits Xij=∑nun(i)vn(j)‾X_{ij}=\sum_nu_n(i)\overline{v_n(j)}Xij​=∑n​un​(i)vn​(j)​ with ∑n∥un∥∥vn∥<∞\sum_n\|u_n\|\|v_n\|<\infty∑n​∥un​∥∥vn​∥<∞; let τ(X)\tau(X)τ(X) be the infimum of these sums, with empty infimum ∞\infty∞. States additionally have real, nonnegative finite-coordinate quadratic forms and summable real diagonal of sum 111, and carry the topology generated by τ\tauτ-balls and its Borel measurable structure. Let B0(X)B_0(X)B0​(X) be a chosen bounded operator with coordinate matrix XXX, or 000 if none exists, and C(T)ij=(Tej)iC(T)_{ij}=(Te_j)_iC(T)ij​=(Tej​)i​. Write rank⁡∞T\operatorname{rank}_\infty Trank∞​T for the algebraic dimension of its range, replacing any infinite value by ∞\infty∞, and ∣E∣∞|E|_\infty∣E∣∞​ for the cardinality of EEE with the same convention. Call XXX full rank when B0(X)x=0B_0(X)x=0B0​(X)x=0 implies x=0x=0x=0. Define the partial trace

(TX)bb′=∑a∈EX(b,a),(b′,a).(TX)_{bb'}=\sum_{a\in E}X_{(b,a),(b',a)}.(TX)bb′​=a∈E∑​X(b,a),(b′,a)​.

For matrices on any relevant index type, define D(X,Y)D(X,Y)D(X,Y) as the supremum over finite coordinate sets sss of

{[Re⁡tr⁡(Xslog⁡fcXs−Xslog⁡fcYs+Ys−Xs)]+,ker⁡Ys⊆ker⁡Xs,∞,otherwise,\begin{cases} [\operatorname{Re}\operatorname{tr}(X_s\log_{\mathrm{fc}}X_s-X_s\log_{\mathrm{fc}}Y_s+Y_s-X_s)]_+, &\ker Y_s\subseteq\ker X_s,\\ \infty,&\text{otherwise}, \end{cases}{[Retr(Xs​logfc​Xs​−Xs​logfc​Ys​+Ys​−Xs​)]+​,∞,​kerYs​⊆kerXs​,otherwise,​

where [a]+=max⁡(a,0)[a]_+=\max(a,0)[a]+​=max(a,0) and log⁡0=0\log0=0log0=0. The first assertion concerns every ensemble e=(wi,ρi)i≥0e=(w_i,\rho_i)_{i\ge0}e=(wi​,ρi​)i≥0​ of states on B×EB\times EB×E, where wi≥0w_i\ge0wi​≥0 and ∑iwi=1\sum_iw_i=1∑i​wi​=1 converges. Its average is the specified reconstructed matrix

ρˉe=C ⁣(B0 ⁣((∑iwiC(B0(ρi))ab)a,b)),\bar\rho_e=C\!\left(B_0\!\left( \left(\sum_iw_iC(B_0(\rho_i))_{ab}\right)_{a,b} \right)\right),ρˉ​e​=C​B0​​(i∑​wi​C(B0​(ρi​))ab​)a,b​​​,

and χF(e)=∑iwiD(F(ρi),F(ρˉe))\chi_F(e)=\sum_iw_iD(F(\rho_i),F(\bar\rho_e))χF​(e)=∑i​wi​D(F(ρi​),F(ρˉ​e​)). If ρˉe\bar\rho_eρˉ​e​ is full rank, there exists r∈Nr\in\mathbb Nr∈N with r<∣E∣∞r<|E|_\inftyr<∣E∣∞​ and rank⁡∞B0(ρi)≤r\operatorname{rank}_\infty B_0(\rho_i)\le rrank∞​B0​(ρi​)≤r for every iii, and χid(e)<∞\chi_{\mathrm{id}}(e)<\inftyχid​(e)<∞, then χT(e)<χid(e)\chi_T(e)<\chi_{\mathrm{id}}(e)χT​(e)<χid​(e). The rank bound applies even at zero-weight indices. The second assertion concerns every Borel probability measure μ\muμ on states of B×EB\times EB×E, with coordinatewise average (ρˉμ)ab=∫ρab dμ(\bar\rho_\mu)_{ab}=\int\rho_{ab}\,d\mu(ρˉ​μ​)ab​=∫ρab​dμ and χF(μ)=∫D(F(ρ),F(ρˉμ)) dμ\chi_F(\mu)=\int D(F(\rho),F(\bar\rho_\mu))\,d\muχF​(μ)=∫D(F(ρ),F(ρˉ​μ​))dμ. If ρˉμ\bar\rho_\muρˉ​μ​ is full rank, there exists r∈Nr\in\mathbb Nr∈N with r<∣E∣∞r<|E|_\inftyr<∣E∣∞​ and rank⁡∞B0(ρ)≤r\operatorname{rank}_\infty B_0(\rho)\le rrank∞​B0​(ρ)≤r for every ρ∈supp⁡μ\rho\in\operatorname{supp}\muρ∈suppμ, and χid(μ)<∞\chi_{\mathrm{id}}(\mu)<\inftyχid​(μ)<∞, then χT(μ)<χid(μ)\chi_T(\mu)<\chi_{\mathrm{id}}(\mu)χT​(μ)<χid​(μ). This rank bound is on the entire topological support. Neither part explicitly requires r>0r>0r>0; if EEE is infinite, its dimensional inequality permits every finite rrr.

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