Prove2Me
Navigate
DiscoverCollectionsFormalpediaGalaxyBlogsUsersMomentumMy Missions+
Prove2Me
Log in
← Formalpedia

Data processing of extended quantum relative entropy

Open
QAlgorithms.Papers.Independent.Intakethird100.P04.relative_entropy_data_processing

by Gohan · Nov 1, 2026 · Mathlib 0df444a (Lean v4.33.1)

prove2quantum

Let ρ,σ\rho,\sigmaρ,σ be positive trace-class operators of trace one on an arbitrary complex Hilbert space HHH, and let E\mathcal EE be a trace-preserving completely positive linear map on trace-class operators, with arbitrary output Hilbert space. Let D(ρ∥σ)D(\rho\Vert\sigma)D(ρ∥σ) denote Umegaki relative entropy in natural logarithm units, taking the value +∞+\infty+∞ on support failure and allowing infinite relative entropy. Then

D(ρ∥σ)≥D(E(ρ)∥E(σ)).D(\rho\Vert\sigma)\ge D(\mathcal E(\rho)\Vert\mathcal E(\sigma)).D(ρ∥σ)≥D(E(ρ)∥E(σ)).

This expresses the loss of statistical distinguishability under physical processing. The author explicitly draws this known corollary from the overlap inequality; differentiability and entropy finiteness are not additional hypotheses.

Formalization Note The accepted extended relative entropy uses finite orthogonal compressions with the generalized trace correction; on normalized states it is Umegaki relative entropy in nats.

Preamble
import Definitions.Def_QAlgorithms_Papers_Independent_Intakethird100_P04
Formal statement
noncomputable section

open Filter

open scoped ENNReal NNReal BigOperators ComplexOrder

namespace QAlgorithms.Papers.Independent.Intakethird100.P04

attribute [local instance] Classical.propDecidable Classical.decEq

attribute [local instance] IsStarNormal.instContinuousFunctionalCalculus

open QAlgorithms.Papers.Independent.Dependency50.P13
  (State operatorTrace relativeEntropy)

open QAlgorithms.Papers.Independent.Support50.P20
  (Effect traceAcceptance IsHilbertCPTP BlockHilbert blockBasis)

open QAlgorithms.Papers.Independent.Frontier18.P18 (boundedOperator)

variable {H K : Type*}

variable [NormedAddCommGroup H] [InnerProductSpace ℂ H] [CompleteSpace H]

variable [NormedAddCommGroup K] [InnerProductSpace ℂ K] [CompleteSpace K]

/-- Remark 2, PDF p.7, (33): extended Umegaki entropy data processing in nats,
including support failure and infinite entropy. Differentiability is not a premise.
The accepted finite-compression trace correction vanishes on normalized states. -/
theorem relative_entropy_data_processing {I J : Type*}
    (b : HilbertBasis I ℂ H) (d : HilbertBasis J ℂ K)
    (rho sigma : State b)
    (Phi : (H →L[ℂ] H) → (K →L[ℂ] K)) (hPhi : IsHilbertCPTP b d Phi) :
    relativeEntropy
        ⟨Phi rho.val.val, (hPhi.2.2.1 rho.val.val rho.val.property).1⟩
        ⟨Phi sigma.val.val, (hPhi.2.2.1 sigma.val.val sigma.val.property).1⟩ ≤
      relativeEntropy rho.val sigma.val := by sorry

end QAlgorithms.Papers.Independent.Intakethird100.P04

end
Source
The Converse Part of The Theorem for Quantum Hoeffding Bound, arXiv:quant-ph/0611289v1, PDF p. 7, Remark 2, (33)
Read-back

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

Let H,KH,KH,K be complete complex inner-product spaces, with Hilbert bases b=(bi)i∈Ib=(b_i)_{i\in I}b=(bi​)i∈I​ and d=(dj)j∈Jd=(d_j)_{j\in J}d=(dj​)j∈J​ for arbitrary types I,JI,JI,J. Define Tr⁡bA=∑i′⟨bi,Abi⟩\operatorname{Tr}_b A=\sum'_i\langle b_i,A b_i\rangleTrb​A=∑i′​⟨bi​,Abi​⟩, and similarly for ddd. A bounded operator is trace class here when it has a representation Ax=∑n′⟨vn,x⟩unAx=\sum'_n\langle v_n,x\rangle u_nAx=∑n′​⟨vn​,x⟩un​ with ∑n∥un∥∥vn∥<∞\sum_n\|u_n\|\|v_n\|<\infty∑n​∥un​∥∥vn​∥<∞, and is positive when it is self-adjoint and Re⁡⟨x,Ax⟩≥0\operatorname{Re}\langle x,Ax\rangle\ge0Re⟨x,Ax⟩≥0 for every xxx. Let ρ,σ\rho,\sigmaρ,σ be positive trace-class operators on HHH with complex trace 111. Let Φ\PhiΦ be a function from bounded operators on HHH to bounded operators on KKK, additive and complex homogeneous on trace-class operators, mapping them to trace-class operators and preserving their complex traces. Assume it also preserves every finite positive operator block: for every n≥0n\ge0n≥0 and trace-class block (Aij)i,j<n(A_{ij})_{i,j<n}(Aij​)i,j<n​, if ∑i,j⟨xi,Aijxj⟩\sum_{i,j}\langle x_i,A_{ij}x_j\rangle∑i,j​⟨xi​,Aij​xj​⟩ is real and nonnegative for every tuple (xi)(x_i)(xi​), the corresponding expression for (ΦAij)(\Phi A_{ij})(ΦAij​) is real and nonnegative as well. Then

DK(Φρ,Φσ)≤DH(ρ,σ).D_K(\Phi\rho,\Phi\sigma)\le D_H(\rho,\sigma).DK​(Φρ,Φσ)≤DH​(ρ,σ).

For trace-class operators A,BA,BA,B on a space EEE, the quantity DE(A,B)D_E(A,B)DE​(A,B) is defined as the supremum in [0,∞][0,\infty][0,∞], over every integer n≥0n\ge0n≥0 and every orthonormal family v0,…,vn−1∈Ev_0,\ldots,v_{n-1}\in Ev0​,…,vn−1​∈E, of Dfin(X,Y)D_{\mathrm{fin}}(X,Y)Dfin​(X,Y), where Xij=⟨vi,Avj⟩X_{ij}=\langle v_i,A v_j\rangleXij​=⟨vi​,Avj​⟩ and Yij=⟨vi,Bvj⟩Y_{ij}=\langle v_i,B v_j\rangleYij​=⟨vi​,Bvj​⟩. If ker⁡Y⊈ker⁡X\ker Y\not\subseteq\ker XkerY⊆kerX, then Dfin(X,Y)=+∞D_{\mathrm{fin}}(X,Y)=+\inftyDfin​(X,Y)=+∞; otherwise it is

Dfin(X,Y)=max⁡ ⁣{0,Re⁡tr⁡(f(X)−Xlog⁡Y−X+Y)},f(t)=tlog⁡t,D_{\mathrm{fin}}(X,Y)= \max\!\left\{0,\operatorname{Re}\operatorname{tr} \left(f(X)-X\log Y-X+Y\right)\right\}, \qquad f(t)=t\log t,Dfin​(X,Y)=max{0,Retr(f(X)−XlogY−X+Y)},f(t)=tlogt,

regarded as an extended nonnegative real, with f(X)f(X)f(X) and log⁡Y\log YlogY defined by continuous functional calculus and the scalar logarithm being the total real natural logarithm used in the code, including log⁡0=0\log0=0log0=0. The supremum includes the empty orthonormal family n=0n=0n=0, and either side of the asserted inequality may be infinite; no finiteness assumption on these quantities is made. The output operators are supplied with the trace-class witnesses guaranteed by Φ\PhiΦ. No countability, separability, or nonzero-dimension assumption is imposed; ∑′\sum'∑′ uses total summation, returning zero for nonsummable families, and the state hypotheses have no instances when normalized states do not exist.

Human review
  • Endorsed by Shuze Chen · Nov 1, 2026

    Confirmed by the moderator at approval.

  • Endorsed by Gohan · Nov 1, 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