Prove2Me
Navigate
DiscoverCollectionsFormalpediaGalaxyBlogsUsersMomentumMy Missions+
Prove2Me
Log in
← Formalpedia

§6.4.2 (6.107): relaxed smooth min/max comparison

Open
QAlgorithms.Papers.Independent.Intakenext100.P07.smooth_relaxed_comparison

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

prove2quantum

All quantum systems below are finite-dimensional complex Hilbert spaces of positive dimension, and logarithms are binary. A normalized state is positive semidefinite with trace one; a subnormalized state has trace at most one. Marginals are partial traces. Here FFF is squared algebraic fidelity, F∗F_*F∗​ is squared generalized fidelity, P=1−F∗P=\sqrt{1-F_*}P=1−F∗​​ is purified distance, and Δ(ρ,τ)=max⁡0≤M≤I∣Tr⁡((ρ−τ)M)∣\Delta(\rho,\tau)=\max_{0\le M\le I}|\operatorname{Tr}((\rho-\tau)M)|Δ(ρ,τ)=max0≤M≤I​∣Tr((ρ−τ)M)∣. The symbols Hmin⁡H_{\min}Hmin​ and Hmax⁡H_{\max}Hmax​ denote conditional min- and max-entropy, with values +∞+\infty+∞ and −∞-\infty−∞ at zero. Smoothing optimizes over all subnormalized states in the purified-distance ball; every entropy radius satisfies 0≤ε<Tr⁡ρ0\le\varepsilon<\sqrt{\operatorname{Tr}\rho}0≤ε<Trρ​.

Let rho_AB be normalized, let epsilon and delta be strictly positive with epsilon+delta<1, and let n>=1. The displayed comparison holds for the actual iid product state; interchanging the two radii gives its symmetric counterpart.

1nHmin⁡ε(An∣Bn)≤1nHmax⁡δ(An∣Bn)+1nlog⁡211−(ε+δ)2.\frac1n H_{\min}^{\varepsilon}(A^n|B^n)\le\frac1n H_{\max}^{\delta}(A^n|B^n)+\frac1n\log_2\frac1{1-(\varepsilon+\delta)^2}.n1​Hminε​(An∣Bn)≤n1​Hmaxδ​(An∣Bn)+n1​log2​1−(ε+δ)21​.

This result records the source’s operational or structural relation, with the stated endpoint and attainment clauses retained.

Preamble
import Definitions.Def_QAlgorithms_Papers_Independent_Intakenext100_P07
Formal statement
noncomputable section

open Filter

open scoped BigOperators Matrix ComplexConjugate ComplexOrder Matrix.Norms.L2Operator Topology

namespace QAlgorithms.Papers.Independent.Intakenext100.P07

attribute [local instance] Classical.propDecidable Classical.decEq

variable {I J A B C D X Y Z : Type*}

variable [Fintype I] [DecidableEq I] [Fintype J] [DecidableEq J]

variable [Fintype A] [DecidableEq A] [Fintype B] [DecidableEq B]

variable [Fintype C] [DecidableEq C] [Fintype D] [DecidableEq D]

variable [Fintype X] [DecidableEq X] [Fintype Y] [DecidableEq Y]

variable [Fintype Z] [DecidableEq Z]

/-- NODE paper:1504.00233#6.relaxedcomparison
§6.4.2 (6.107): relaxed smooth min/max comparison, arXiv:1504.00233v5, PDF p.110. -/
theorem smooth_relaxed_comparison [Nonempty A] [Nonempty B] (r : paperMat (A × B)) (hr : IsDensityMatrix r)
    (e d : ℝ) (he : 0 < e) (hd : 0 < d) (hed : e+d < 1) (n : ℕ) (hn : 0 < n) :
    minRate r e n ≤ maxRate r d n+((Real.logb 2 (1/(1-(e+d)^2))/(n:ℝ):ℝ):EReal) ∧
    minRate r d n ≤ maxRate r e n+((Real.logb 2 (1/(1-(e+d)^2))/(n:ℝ):ℝ):EReal) := by sorry

end QAlgorithms.Papers.Independent.Intakenext100.P07

end
Source
Quantum Information Processing with Finite Resources -- Mathematical Foundations, arXiv:1504.00233v5, PDF p. 110, §6.4.2 (6.107): relaxed smooth min/max comparison
Read-back

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

Let A,BA,BA,B be nonempty finite types with decidable equality, and let rrr be a positive semidefinite complex matrix on A×BA\times BA×B with trace 111. Let e,d>0e,d>0e,d>0 satisfy e+d<1e+d<1e+d<1, and let nnn be a positive integer. Put r(x,y),(x′,y′)[n]=∏i=0n−1r(xi,yi),(xi′,yi′)r^{[n]}_{(x,y),(x',y')}=\prod_{i=0}^{n-1}r_{(x_i,y_i),(x'_i,y'_i)}r(x,y),(x′,y′)[n]​=∏i=0n−1​r(xi​,yi​),(xi′​,yi′​)​, indexed by An×BnA^n\times B^nAn×Bn. For matrices on any such bipartite space U×VU\times VU×V, define R(u,v)=Re⁡tr⁡u vuR(u,v)=\operatorname{Re}\operatorname{tr}\sqrt{\sqrt u\,v\sqrt u}R(u,v)=Retru​vu​​ by real functional calculus, P(u,v)=1−(R(u,v)+(1−Re⁡tr⁡u)(1−Re⁡tr⁡v))2P(u,v)=\sqrt{1-(R(u,v)+\sqrt{(1-\operatorname{Re}\operatorname{tr}u)(1-\operatorname{Re}\operatorname{tr}v)})^2}P(u,v)=1−(R(u,v)+(1−Retru)(1−Retrv)​)2​, and Ba(v)={u:u⪰0, Re⁡tr⁡u≤1, P(u,v)≤a}\mathcal B_a(v)=\{u:u\succeq0,\ \operatorname{Re}\operatorname{tr}u\le1,\ P(u,v)\le a\}Ba​(v)={u:u⪰0, Retru≤1, P(u,v)≤a}. Define Hmin⁡(u)=sup⁡{ℓ∈R:∃s⪰0 on V, Re⁡tr⁡s≤1, u⪯2−ℓ(IU⊗s)}H_{\min}(u)=\sup\{\ell\in\mathbb R:\exists s\succeq0\text{ on }V,\ \operatorname{Re}\operatorname{tr}s\le1,\ u\preceq2^{-\ell}(I_U\otimes s)\}Hmin​(u)=sup{ℓ∈R:∃s⪰0 on V, Retrs≤1, u⪯2−ℓ(IU​⊗s)}, Hmax⁡(u)=sup⁡s⪰0, Re⁡tr⁡s≤1L(R(u,IU⊗s)2)H_{\max}(u)=\sup_{s\succeq0,\ \operatorname{Re}\operatorname{tr}s\le1}L(R(u,I_U\otimes s)^2)Hmax​(u)=sups⪰0, Retrs≤1​L(R(u,IU​⊗s)2), Hmin⁡a(v)=sup⁡u∈Ba(v)Hmin⁡(u)H_{\min}^a(v)=\sup_{u\in\mathcal B_a(v)}H_{\min}(u)Hmina​(v)=supu∈Ba​(v)​Hmin​(u), and Hmax⁡a(v)=inf⁡u∈Ba(v)Hmax⁡(u)H_{\max}^a(v)=\inf_{u\in\mathcal B_a(v)}H_{\max}(u)Hmaxa​(v)=infu∈Ba​(v)​Hmax​(u), where all entropy extrema are extended-real, L(0)=−∞L(0)=-\inftyL(0)=−∞, L(x)=log⁡2xL(x)=\log_2xL(x)=log2​x otherwise, and ⪯\preceq⪯ denotes positive semidefinite difference. Then both inequalities hold:

1nHmin⁡e(r[n])≤1nHmax⁡d(r[n])+log⁡2 ⁣(1/(1−(e+d)2))n,1nHmin⁡d(r[n])≤1nHmax⁡e(r[n])+log⁡2 ⁣(1/(1−(e+d)2))n.\frac1nH_{\min}^e(r^{[n]}) \le \frac1nH_{\max}^d(r^{[n]}) +\frac{\log_2\!\left(1/(1-(e+d)^2)\right)}n, \qquad \frac1nH_{\min}^d(r^{[n]}) \le \frac1nH_{\max}^e(r^{[n]}) +\frac{\log_2\!\left(1/(1-(e+d)^2)\right)}n.n1​Hmine​(r[n])≤n1​Hmaxd​(r[n])+nlog2​(1/(1−(e+d)2))​,n1​Hmind​(r[n])≤n1​Hmaxe​(r[n])+nlog2​(1/(1−(e+d)2))​.

The factors 1/n1/n1/n and the added real quantities are embedded in the extended reals, where the arithmetic is performed.

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