Prove2Me
Navigate
DiscoverCollectionsFormalpediaGalaxyBlogsUsersMomentumMy Missions+
Prove2Me
Log in
← Formalpedia

Proposition 4.5 / (54)–(57): large subspace with high minimum fidelity

Open
QAlgorithms.Papers.Independent.Gap20.P07.largeSubspace_fidelity

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

prove2quantum

Let TTT be a CPTP endomorphism of a complex system of dimension d≥1d\ge1d≥1, let ρ\rhoρ be a density matrix, and let 1≤k≤d1\le k\le d1≤k≤d. There is an isometry V:Ck→CdV:\mathbb C^k\to\mathbb C^dV:Ck→Cd, with P=VV∗P=VV^*P=VV∗, such that the compressed CPTP channel is

Tk(X)=V∗T(VXV∗)V+tr⁡((I−P)T(VXV∗))kIk.T_k(X)=V^*T(VXV^*)V+\frac{\operatorname{tr}((I-P)T(VXV^*))}{k}I_k.Tk​(X)=V∗T(VXV∗)V+ktr((I−P)T(VXV∗))​Ik​.

For this same subspace the minimum normalized pure-state fidelity satisfies

Fmin⁡(Tk)≥1−1−Fe(ρ,T)1−qF_{\min}(T_k)\ge1-\frac{1-F_e(\rho,T)}{1-q}Fmin​(Tk​)≥1−1−q1−Fe​(ρ,T)​

with q=k∥ρ∥∞q=k\|\rho\|_\inftyq=k∥ρ∥∞​ whenever this q<1q<1q<1, and also with q=(1+log⁡2d−S(ρ))/(log⁡2d−log⁡2k)q=(1+\log_2d-S(\rho))/(\log_2d-\log_2k)q=(1+log2​d−S(ρ))/(log2​d−log2​k) whenever k<dk<dk<d and this second q<1q<1q<1. The guards are independent. Entropy is in bits and FeF_eFe​ is squared. The guarded estimates supply the subspace selection needed for capacity equivalences; unguarded printed denominators are corrected explicitly.

Preamble
import Definitions.Def_QAlgorithms_Papers_Independent_Gap20_P07
Formal statement
noncomputable section

open Filter MeasureTheory

open scoped Topology ENNReal NNReal Matrix ComplexOrder ComplexConjugate

attribute [local instance] Classical.propDecidable

attribute [local instance] Classical.decEq

attribute [local instance] IsStarNormal.instContinuousFunctionalCalculus

namespace QAlgorithms.Papers.Independent.Gap20.P07

open QAlgorithms

open QAlgorithms.RealRegisters (Hilbert)

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

open QAlgorithms.Papers.Independent.Frontier30.P14
  (Operation isometryCoeff conjugateKernel)

open QAlgorithms.Papers.Independent.Claimed50.P35
  (stabilizedTraceClassNorm traceClassTensorMap traceClassChannelPower
   transpose transposeBound TraceLinear tracePair MeasurePrepare)

open QAlgorithms.Papers.Independent.Gap48.P29 (fe FiniteCPMap)

/-- PDF pp.17--18, Proposition 4.5, (54)--(57). Printed denominators lacked
necessary guards. Changed: 1<=k<=d, with each q<1 guard independently applied,
and k<d for the entropy quotient. The same chosen range satisfies both estimates.
The half-space application of (55) is 1-2(1-Fe), correcting p.18's reciprocal typo. -/
theorem largeSubspace_fidelity (d k : ℕ) (hd : 0 < d) (hk : 0 < k) (hkd : k ≤ d)
    (T : Superop (Fin d) (Fin d)) (hT : IsChannel T)
    (rho : Matrix (Fin d) (Fin d) ℂ) (hrho : IsDensityMatrix rho) :
    ∃ V : Matrix (Fin d) (Fin k) ℂ, V.conjTranspose * V = 1 ∧
      IsChannel (compressedChannel V T) ∧
      ((k : ℝ) * specNorm rho < 1 →
        1 - (1 - fe rho T) / (1 - (k : ℝ) * specNorm rho) ≤
          minimumFidelity k (compressedChannel V T)) ∧
      (k < d → let q := (1 + Real.logb 2 d - vnEntropy rho) /
        (Real.logb 2 d - Real.logb 2 k)
        q < 1 → 1 - (1 - fe rho T) / (1 - q) ≤ minimumFidelity k (compressedChannel V T)) := by sorry

end QAlgorithms.Papers.Independent.Gap20.P07

end
Source
Tema Con Variazioni: Quantum Channel Capacity, arXiv:quant-ph/0311037v1, PDF p. 17, Proposition 4.5 / (54)–(57): large subspace with high minimum fidelity
Read-back

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

For all positive integers d,kd,kd,k with k≤dk\le dk≤d, every map T:Md(C)→Md(C)T:M_d(\mathbb C)\to M_d(\mathbb C)T:Md​(C)→Md​(C) admitting a finite Kraus representation T(X)=∑aKaXKa†T(X)=\sum_aK_aXK_a^\daggerT(X)=∑a​Ka​XKa†​ with ∑aKa†Ka=Id\sum_aK_a^\dagger K_a=I_d∑a​Ka†​Ka​=Id​, and every positive semidefinite ρ∈Md(C)\rho\in M_d(\mathbb C)ρ∈Md​(C) of trace 111, there exists a matrix V∈Md×k(C)V\in M_{d\times k}(\mathbb C)V∈Md×k​(C) with V†V=IkV^\dagger V=I_kV†V=Ik​ such that the map

CV(X)=V†T(VXV†)V+tr⁡ ⁣((Id−VV†)T(VXV†))k IkC_V(X)=V^\dagger T(VXV^\dagger)V+ \frac{\operatorname{tr}\!\bigl((I_d-VV^\dagger)T(VXV^\dagger)\bigr)}{k}\,I_kCV​(X)=V†T(VXV†)V+ktr((Id​−VV†)T(VXV†))​Ik​

also admits a finite Kraus representation satisfying the corresponding identity sum, and this same VVV satisfies both following conditional bounds. Define

Fe(ρ,T)=Re⁡∑i,j,r,s=0d−1ρjiρsr‾ [T(Ejs)]ir,fmin⁡(CV)=inf⁡ψ∈Ck∥ψ∥2=1Re⁡tr⁡ ⁣((ψψ†)CV(ψψ†)),F_e(\rho,T)= \operatorname{Re}\sum_{i,j,r,s=0}^{d-1} \rho_{ji}\overline{\rho_{sr}}\,[T(E_{js})]_{ir}, \qquad f_{\min}(C_V)=\inf_{\substack{\psi\in\mathbb C^k\\\|\psi\|_2=1}} \operatorname{Re}\operatorname{tr}\!\bigl((\psi\psi^\dagger)C_V(\psi\psi^\dagger)\bigr),Fe​(ρ,T)=Rei,j,r,s=0∑d−1​ρji​ρsr​​[T(Ejs​)]ir​,fmin​(CV​)=ψ∈Ck∥ψ∥2​=1​inf​Retr((ψψ†)CV​(ψψ†)),

where EjsE_{js}Ejs​ is a matrix unit, and let S(ρ)=−Re⁡tr⁡(cfc[x↦xlog⁡2x](ρ))S(\rho)=-\operatorname{Re}\operatorname{tr}(cfc[x\mapsto x\log_2x](\rho))S(ρ)=−Retr(cfc[x↦xlog2​x](ρ)), with cfccfccfc denoting continuous functional calculus. If k∥ρ∥op<1k\|\rho\|_{\mathrm{op}}<1k∥ρ∥op​<1, then

1−1−Fe(ρ,T)1−k∥ρ∥op≤fmin⁡(CV).1-\frac{1-F_e(\rho,T)}{1-k\|\rho\|_{\mathrm{op}}}\le f_{\min}(C_V).1−1−k∥ρ∥op​1−Fe​(ρ,T)​≤fmin​(CV​).

If k<dk<dk<d, put

q=1+log⁡2d−S(ρ)log⁡2d−log⁡2k;q=\frac{1+\log_2d-S(\rho)}{\log_2d-\log_2k};q=log2​d−log2​k1+log2​d−S(ρ)​;

whenever q<1q<1q<1, one has

1−1−Fe(ρ,T)1−q≤fmin⁡(CV).1-\frac{1-F_e(\rho,T)}{1-q}\le f_{\min}(C_V).1−1−q1−Fe​(ρ,T)​≤fmin​(CV​).

The case k=dk=dk=d is included, with the second bound conditional on the unsatisfied inequality k<dk<dk<d.

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