Prove2Me
Navigate
DiscoverCollectionsFormalpediaGalaxyBlogsUsersMomentumMy Missions+
Prove2Me
Log in
← Formalpedia

Properties of the sharp entropy-difference bound

Open
QAlgorithms.Papers.Independent.Support50.P15.theorem2

by Goku · Oct 23, 2026 · Mathlib 0df444a (Lean v4.33.1)

prove2quantum

For finite d≥2d\ge2d≥2, let M(Δ,d)M(\Delta,d)M(Δ,d) be the constrained binary-relative-entropy minimum of Eq. (6), N(d)=max⁡0<r<1/2r(1−r)[log⁡(((1−r)/r)(d−1))]2N(d)=\max_{0<r<1/2}r(1-r)[\log(((1-r)/r)(d-1))]^2N(d)=max0<r<1/2​r(1−r)[log(((1−r)/r)(d−1))]2 and Nd=14log⁡2(d−1)+1N_d=\tfrac14\log^2(d-1)+1Nd​=41​log2(d−1)+1. The function MMM is nonnegative, continuous in the extended order topology and strictly convex on [−log⁡d,log⁡d][-\log d,\log d][−logd,logd], and its finite restriction is continuously differentiable in the interior. Its endpoint values are

M(0,d)=0,M(−log⁡d,d)=log⁡d,M(log⁡d,d)=+∞.M(0,d)=0,\quad M(-\log d,d)=\log d,\quad M(\log d,d)=+\infty.M(0,d)=0,M(−logd,d)=logd,M(logd,d)=+∞.

It is finite on [−log⁡d,log⁡d)[-\log d,\log d)[−logd,logd). One has 0<N(d)<Nd0<N(d)<N_d0<N(d)<Nd​ and N(d)<log⁡2dN(d)<\log^2 dN(d)<log2d. For every real N≥N(d)N\ge N(d)N≥N(d) and every real Δ\DeltaΔ, using the empty-feasible-set extension of MMM,

M(Δ,d)≥N(eΔ/N−1−Δ/N)≥Δ22N+Δ36N2.M(\Delta,d)\ge N(e^{\Delta/N}-1-\Delta/N)\ge\frac{\Delta^2}{2N}+\frac{\Delta^3}{6N^2}.M(Δ,d)≥N(eΔ/N−1−Δ/N)≥2NΔ2​+6N2Δ3​.

On the entropy interval one also has M(Δ,d)≥Δ2/(3log⁡2d)M(\Delta,d)\ge\Delta^2/(3\log^2 d)M(Δ,d)≥Δ2/(3log2d). These provide explicit lower bounds without optimizing the scalar pair.

Preamble
import Definitions.Def_QAlgorithms_Papers_Independent_Support50_P15
Formal statement
noncomputable section

attribute [local instance] Classical.propDecidable

open scoped Matrix ComplexConjugate ENNReal Matrix.Norms.L2Operator

open Filter MeasureTheory

namespace QAlgorithms.Papers.Independent.Support50.P15

open QAlgorithms.Papers.Independent.Information100.P56 (entropy)

/-- Theorem 2, PDF p.4; empty-feasible-set extension of (11) on p.21.
Continuity is extended-real continuity; differentiability uses only the finite interior. -/
theorem theorem2 (d : ℕ) (hd : 2 ≤ d) :
    (∀ delta ∈ Set.Icc (-Real.log d) (Real.log d), 0 ≤ tightBound d delta) ∧
    ContinuousOn (tightBound d) (Set.Icc (-Real.log d) (Real.log d)) ∧
    (∀ x ∈ Set.Icc (-Real.log d) (Real.log d),
      ∀ y ∈ Set.Icc (-Real.log d) (Real.log d), x ≠ y →
      ∀ t ∈ Set.Ioo (0 : ℝ) 1,
        tightBound d ((1 - t) * x + t * y) <
          ((1 - t : ℝ) : EReal) * tightBound d x + (t : EReal) * tightBound d y) ∧
    ContDiffOn ℝ 1 (fun x => (tightBound d x).toReal) (Set.Ioo (-Real.log d) (Real.log d)) ∧
    tightBound d 0 = 0 ∧ tightBound d (-Real.log d) = (Real.log d : ℝ) ∧
    tightBound d (Real.log d) = ⊤ ∧
    (∀ delta ∈ Set.Ico (-Real.log d) (Real.log d), tightBound d delta < ⊤) ∧
    0 < maxVariance d ∧ maxVariance d < easyVariance d ∧ maxVariance d < (Real.log d) ^ 2 ∧
    (∀ n : ℝ, maxVariance d ≤ n → ∀ delta : ℝ,
      (exponentialBound n delta : EReal) ≤ tightBound d delta ∧
      cubicBound n delta ≤ exponentialBound n delta) ∧
    (∀ delta ∈ Set.Icc (-Real.log d) (Real.log d),
      (delta ^ 2 / (3 * (Real.log d) ^ 2) : ℝ) ≤ tightBound d delta) := by sorry

end QAlgorithms.Papers.Independent.Support50.P15

end
Source
Tight bound on relative entropy by entropy difference, arXiv:1304.0036v3, PDF p. 4, theorem2
Read-back

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

For every natural number d≥2d\ge2d≥2, define

Hd(u)=−ulog⁡u−(1−u)log⁡(1−u)+ulog⁡(d−1),Bd(δ)=inf⁡0≤u,v≤(d−1)/dHd(u)−Hd(v)=δK(u,v),H_d(u)=-u\log u-(1-u)\log(1-u)+u\log(d-1),\qquad B_d(\delta)=\inf_{\substack{0\le u,v\le(d-1)/d\\H_d(u)-H_d(v)=\delta}}K(u,v),Hd​(u)=−ulogu−(1−u)log(1−u)+ulog(d−1),Bd​(δ)=0≤u,v≤(d−1)/dHd​(u)−Hd​(v)=δ​inf​K(u,v),

where K(u,v)=k(u,v)+k(1−u,1−v)K(u,v)=k(u,v)+k(1-u,1-v)K(u,v)=k(u,v)+k(1−u,1−v), k(p,q)=0k(p,q)=0k(p,q)=0 when p=0p=0p=0, k(p,q)=+∞k(p,q)=+\inftyk(p,q)=+∞ when p≠0p\ne0p=0 and q=0q=0q=0, and k(p,q)=plog⁡(p/q)k(p,q)=p\log(p/q)k(p,q)=plog(p/q) otherwise; the infimum is extended-real and is +∞+\infty+∞ for an empty feasible set. All logarithms are natural and log⁡0=0\log0=0log0=0. Also put

Vd=sup⁡0<r<1/2r(1−r)[log⁡ ⁣(1−rr(d−1))]2,Ed=log⁡2(d−1)4+1,V_d=\sup_{0<r<1/2}r(1-r)\left[\log\!\left(\frac{1-r}{r}(d-1)\right)\right]^2, \quad E_d=\frac{\log^2(d-1)}4+1,Vd​=0<r<1/2sup​r(1−r)[log(r1−r​(d−1))]2,Ed​=4log2(d−1)​+1, L(n,δ)=n(eδ/n−1−δ/n),C(n,δ)=δ22n+δ36n2.L(n,\delta)=n\bigl(e^{\delta/n}-1-\delta/n\bigr), \qquad C(n,\delta)=\frac{\delta^2}{2n}+\frac{\delta^3}{6n^2}.L(n,δ)=n(eδ/n−1−δ/n),C(n,δ)=2nδ2​+6n2δ3​.

The theorem asserts that BdB_dBd​ is nonnegative and continuous on [−log⁡d,log⁡d][-\log d,\log d][−logd,logd], with continuity understood in the extended-real topology; for every distinct x,yx,yx,y in that interval and every 0<t<10<t<10<t<1,

Bd((1−t)x+ty)<(1−t)Bd(x)+tBd(y),B_d((1-t)x+ty)<(1-t)B_d(x)+tB_d(y),Bd​((1−t)x+ty)<(1−t)Bd​(x)+tBd​(y),

using extended-real arithmetic. It further asserts that the real-valued function obtained by converting BdB_dBd​ to a real number is continuously differentiable on (−log⁡d,log⁡d)(-\log d,\log d)(−logd,logd); this conversion assigns 000 to either infinity. The specified values and finiteness are

Bd(0)=0,Bd(−log⁡d)=log⁡d,Bd(log⁡d)=+∞,Bd(δ)<+∞ for −log⁡d≤δ<log⁡d.B_d(0)=0,\qquad B_d(-\log d)=\log d,\qquad B_d(\log d)=+\infty, \qquad B_d(\delta)<+\infty\ \text{for }-\log d\le\delta<\log d.Bd​(0)=0,Bd​(−logd)=logd,Bd​(logd)=+∞,Bd​(δ)<+∞ for −logd≤δ<logd.

Finally,

0<Vd<Ed,Vd<(log⁡d)2;0<V_d<E_d,\qquad V_d<(\log d)^2;0<Vd​<Ed​,Vd​<(logd)2;

for every real n≥Vdn\ge V_dn≥Vd​ and every real δ\deltaδ, both L(n,δ)≤Bd(δ)L(n,\delta)\le B_d(\delta)L(n,δ)≤Bd​(δ) and C(n,δ)≤L(n,δ)C(n,\delta)\le L(n,\delta)C(n,δ)≤L(n,δ) hold; and for every δ∈[−log⁡d,log⁡d]\delta\in[-\log d,\log d]δ∈[−logd,logd],

δ23(log⁡d)2≤Bd(δ).\frac{\delta^2}{3(\log d)^2}\le B_d(\delta).3(logd)2δ2​≤Bd​(δ).
Human review
  • Endorsed by Shuze Chen · Oct 24, 2026

    Confirmed by the moderator at approval.

  • Endorsed by Goku · Oct 24, 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