Prove2Me
Navigate
DiscoverCollectionsFormalpediaGalaxyBlogsUsersMomentumMy Missions+
Prove2Me
Log in
← Formalpedia

Remark 6.2

Open
QAlgorithms.Papers.Independent.Support50.P12.remark_6_2

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

prove2quantum

Let A,BA,BA,B be positive semidefinite operators on a nonzero finite-dimensional complex Hilbert space. Let f:[0,∞)→Rf:[0,\infty)\to\mathbb Rf:[0,∞)→R be continuous on (0,∞)(0,\infty)(0,∞) with signed extended recession slope ω(f)=lim⁡x→∞f(x)/x\omega(f)=\lim_{x\to\infty}f(x)/xω(f)=limx→∞​f(x)/x. Write SfS_fSf​ for the paper's quantum divergence, Qf(A,B)=⟨B1/2,f(LARB−1)B1/2⟩HSQ_f(A,B)=\langle B^{1/2},f(L_AR_{B^{-1}})B^{1/2}\rangle_{\rm HS}Qf​(A,B)=⟨B1/2,f(LA​RB−1​)B1/2⟩HS​, and X0X^0X0 for the support projection. All powers vanish on the kernel, including exponent zero. Traces are representation traces. Write Sα(A∥B)S_\alpha(A\Vert B)Sα​(A∥B) for the unnormalized Petz Rényi entropy, ψA,B(α)=log⁡Tr⁡AαB1−α\psi_{A,B}(\alpha)=\log\operatorname{Tr}A^\alpha B^{1-\alpha}ψA,B​(α)=logTrAαB1−α, C(A∥B)=sup⁡0≤α<1−ψA,B(α)C(A\Vert B)=\sup_{0\le\alpha<1}-\psi_{A,B}(\alpha)C(A∥B)=sup0≤α<1​−ψA,B​(α), and Hr(A∥B)=sup⁡0≤α<1[−αr−ψA,B(α)]/(1−α)H_r(A\Vert B)=\sup_{0\le\alpha<1}[-\alpha r-\psi_{A,B}(\alpha)]/(1-\alpha)Hr​(A∥B)=sup0≤α<1​[−αr−ψA,B​(α)]/(1−α). Logarithms are natural and log⁡0=−∞\log0=-\inftylog0=−∞. The Petz map is ΦB∗(Y)=B1/2Φ∗(Φ(B)−1/2YΦ(B)−1/2)B1/2\Phi_B^*(Y)=B^{1/2}\Phi^*(\Phi(B)^{-1/2}Y\Phi(B)^{-1/2})B^{1/2}ΦB∗​(Y)=B1/2Φ∗(Φ(B)−1/2YΦ(B)−1/2)B1/2.

Put ψ~(s)=(1+s)ψ(s/(1+s))\widetilde\psi(s)=(1+s)\psi(s/(1+s))ψ​(s)=(1+s)ψ(s/(1+s)) for s≥0s\ge0s≥0 and +∞+\infty+∞ for s<0s<0s<0, and h∗(r)=sup⁡s{sr−h(s)}h^*(r)=\sup_s\{sr-h(s)\}h∗(r)=sups​{sr−h(s)}. Then Hr=ψ~∗(−r)H_r=\widetilde\psi^*(-r)Hr​=ψ​∗(−r) is convex and decreasing in rrr, ψ(0)=−S0(A∥B)\psi(0)=-S_0(A\Vert B)ψ(0)=−S0​(A∥B), ψ(1)=−S0(B∥A)\psi(1)=-S_0(B\Vert A)ψ(1)=−S0​(B∥A), and

−Sα(A∥B)=sup⁡r∈R{−rα1−α−Hr(A∥B)},0≤α<1.-S_\alpha(A\Vert B)=\sup_{r\in\mathbb R}\left\{-\frac{r\alpha}{1-\alpha}-H_r(A\Vert B)\right\},\quad0\le\alpha<1.−Sα​(A∥B)=r∈Rsup​{−1−αrα​−Hr​(A∥B)},0≤α<1.

For nonorthogonal pairs, ψ\psiψ is real convex and ψ~\widetilde\psiψ​ is lower semicontinuous and convex. Its derivative is ψ(s/(1+s))+ψ′(s/(1+s))/(1+s)\psi(s/(1+s))+\psi\prime(s/(1+s))/(1+s)ψ(s/(1+s))+ψ′(s/(1+s))/(1+s), its right derivative at zero is ψ(0)+ψ′(0)\psi(0)+\psi\prime(0)ψ(0)+ψ′(0), and its derivative tends to ψ(1)\psi(1)ψ(1) at infinity. If −r<ψ(0)+ψ′(0)-r<\psi(0)+\psi\prime(0)−r<ψ(0)+ψ′(0), Hr=−ψ(0)H_r=-\psi(0)Hr​=−ψ(0); if −r>ψ(1)-r>\psi(1)−r>ψ(1), Hr=+∞H_r=+\inftyHr​=+∞. Bipolarity holds on s≥0s\ge0s≥0, and the displayed supremum can be restricted to the closed interval [−ψ(1),−ψ(0)−ψ′(0)][-\psi(1),-\psi(0)-\psi\prime(0)][−ψ(1),−ψ(0)−ψ′(0)], including a singleton. If Tr⁡A≤1\operatorname{Tr}A\le1TrA≤1, SαS_\alphaSα​ increases on [0,1)[0,1)[0,1), lim⁡α↑1Sα=H0=:S1\lim_{\alpha\uparrow1}S_\alpha=H_0=:S_1limα↑1​Sα​=H0​=:S1​, and lim⁡r↓0Hr=H0\lim_{r\downarrow0}H_r=H_0limr↓0​Hr​=H0​. When Tr⁡A=1\operatorname{Tr}A=1TrA=1, S1=S(A∥B)S_1=S(A\Vert B)S1​=S(A∥B). No derivative is assigned to the identically −∞-\infty−∞ orthogonal case. These identities relate the Rényi and Hoeffding families.

Preamble
import Definitions.Def_QAlgorithms_Papers_Independent_Support50_P12
Formal statement
noncomputable section

namespace QAlgorithms.Papers.Independent.Support50.P12

open Filter MeasureTheory Set

open QAlgorithms.Papers.Independent.Information100.P60 (Mat copies)

open QAlgorithms.Papers.Independent.Information100.P30 (MatrixMap)

open QAlgorithms.Papers.Independent.Dependency50.P28 (tr LE logNonnegativeExtended pinch)

open QAlgorithms.Papers.Independent.Dependency50.P21 (channelAdjoint spectralFunction petz relativeEntropy)

open QAlgorithms.Papers.Independent.Frontier18.P17 (hs)

open QAlgorithms.Papers.Independent.Next46.P23 (Supports)

open QAlgorithms.Papers.Independent.Support21.P03 (ExtendedConvex)

open QAlgorithms (spectralProjector)

open QAlgorithms.Learning (IsDistribution)

open QAlgorithms.Papers.Independent.Information100.P04 (absOp)

open scoped BigOperators Matrix ComplexConjugate ComplexOrder ENNReal Topology Kronecker

attribute [local instance] Classical.propDecidable Classical.decEq

open QAlgorithms.Papers.Independent.Information100.P71 (supportPower)

variable {I J K : Type} [Fintype I] [DecidableEq I] [Fintype J] [DecidableEq J]
  [Fintype K] [DecidableEq K]

/-- Remark 6.2, PDF pp.27–28: polar duality, restricted CLOSED interval, derivatives
on nonorthogonal pairs, and the trace-guarded order-one and right-rate limits.
The separately deferred unnormalized endpoint-entropy derivative formulas are omitted. -/
theorem remark_6_2 [Nonempty I] (X Y : Mat I) (hX : X.PosSemidef) (hY : Y.PosSemidef) :
    (∀ r : ℝ, hoeffding X Y r = polar (logOverlapPerspective X Y) (-r)) ∧
    (∀ r s : ℝ, r ≤ s → hoeffding X Y s ≤ hoeffding X Y r) ∧
    ExtendedConvex (hoeffding X Y) ∧
    logOverlap X Y 0 = -petzRenyi 0 X Y ∧ logOverlap X Y 1 = -petzRenyi 0 Y X ∧
    (∀ a ∈ Ico (0 : ℝ) 1, -petzRenyi a X Y =
      sSup {v | ∃ r : ℝ, v = ((-r * a / (1 - a) : ℝ) : EReal) - hoeffding X Y r}) ∧
    (0 < tr (supp X * supp Y) →
      ConvexOn ℝ univ (psiReal X Y) ∧ ExtendedConvex (logOverlapPerspective X Y) ∧
      LowerSemicontinuous (logOverlapPerspective X Y) ∧
      (∀ s > 0, HasDerivAt (fun u => (logOverlapPerspective X Y u).toReal)
        (psiReal X Y (s / (1 + s)) + deriv (psiReal X Y) (s / (1 + s)) / (1 + s)) s) ∧
      HasDerivWithinAt (fun u => (logOverlapPerspective X Y u).toReal)
        (psiReal X Y 0 + deriv (psiReal X Y) 0) (Ici 0) 0 ∧
      Tendsto (fun s : ℝ => deriv (fun u => (logOverlapPerspective X Y u).toReal) s)
        atTop (𝓝 (psiReal X Y 1)) ∧
      (∀ r : ℝ, -r < psiReal X Y 0 + deriv (psiReal X Y) 0 →
        hoeffding X Y r = -logOverlap X Y 0) ∧
      (∀ r : ℝ, psiReal X Y 1 < -r → hoeffding X Y r = ⊤) ∧
      (∀ s ≥ 0, logOverlapPerspective X Y s = polar (polar (logOverlapPerspective X Y)) s) ∧
      (∀ a ∈ Ico (0 : ℝ) 1, -petzRenyi a X Y =
        sSup {v | ∃ r : ℝ, -psiReal X Y 1 ≤ r ∧
          r ≤ -psiReal X Y 0 - deriv (psiReal X Y) 0 ∧
          v = ((-r * a / (1 - a) : ℝ) : EReal) - hoeffding X Y r})) ∧
    (tr X ≤ 1 →
      MonotoneOn (fun a => petzRenyi a X Y) (Ico 0 1) ∧
      Tendsto (fun a => petzRenyi a X Y) (𝓝[<] 1) (𝓝 (hoeffding X Y 0)) ∧
      Tendsto (hoeffding X Y) (𝓝[>] 0) (𝓝 (hoeffding X Y 0)) ∧
      hoeffding X Y 0 = renyiOne X Y) ∧
    (tr X = 1 → renyiOne X Y = relativeEntropy X Y) := by sorry

end QAlgorithms.Papers.Independent.Support50.P12

end
Source
Quantum f-divergences and error correction, arXiv:1008.2529v6, PDF p. 27, Remark 6.2
Read-back

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

Let III be a finite, nonempty index type with decidable equality, and let X,YX,YX,Y be positive semidefinite complex matrices indexed by III, without any trace normalization assumption. Write tr⁡Z=Re⁡Tr⁡Z\operatorname{tr}Z=\operatorname{Re}\operatorname{Tr}ZtrZ=ReTrZ, and write Z[a]Z^{[a]}Z[a] for spectral functional calculus with the function that is xax^axa when x≠0x\ne0x=0 and 000 when x=0x=0x=0; in particular, SZ=Z[0]S_Z=Z^{[0]}SZ​=Z[0] is the support projection, including when Z=0Z=0Z=0. Define ℓ(q)=ln⁡q\ell(q)=\ln qℓ(q)=lnq for q>0q>0q>0 and ℓ(q)=−∞\ell(q)=-\inftyℓ(q)=−∞ for q≤0q\le0q≤0, and set

L(a)=ℓ ⁣(tr⁡(X[a]Y[1−a])),ψ(a)=L(a).toReal,Da(X,Y)=1a−1L(a)(a<1),L(a)=\ell\!\left(\operatorname{tr}\bigl(X^{[a]}Y^{[1-a]}\bigr)\right),\qquad \psi(a)=L(a).\mathrm{toReal},\qquad D_a(X,Y)=\frac{1}{a-1}L(a)\quad(a<1),L(a)=ℓ(tr(X[a]Y[1−a])),ψ(a)=L(a).toReal,Da​(X,Y)=a−11​L(a)(a<1),

where toReal\mathrm{toReal}toReal sends either infinite extended-real value to 000, and all extended-real arithmetic uses its total conventions. Define

H(r)=sup⁡0≤a<1−ar−L(a)1−a,P(s)={(1+s)L ⁣(s1+s),s≥0,+∞,s<0,f∗(r)=sup⁡s∈R(sr−f(s)).H(r)=\sup_{0\le a<1}\frac{-ar-L(a)}{1-a},\qquad P(s)= \begin{cases} (1+s)L\!\left(\frac{s}{1+s}\right),&s\ge0,\\ +\infty,&s<0, \end{cases} \qquad f^*(r)=\sup_{s\in\mathbb R}\bigl(sr-f(s)\bigr).H(r)=0≤a<1sup​1−a−ar−L(a)​,P(s)={(1+s)L(1+ss​),+∞,​s≥0,s<0,​f∗(r)=s∈Rsup​(sr−f(s)).

Then H(r)=P∗(−r)H(r)=P^*(-r)H(r)=P∗(−r) for every real rrr; H(s)≤H(r)H(s)\le H(r)H(s)≤H(r) whenever r≤sr\le sr≤s; and HHH is extended convex in the precise sense that, whenever H(a)≤uH(a)\le uH(a)≤u and H(b)≤vH(b)\le vH(b)≤v for real a,b,u,va,b,u,va,b,u,v, one has H(wa+(1−w)b)≤wu+(1−w)vH(wa+(1-w)b)\le wu+(1-w)vH(wa+(1−w)b)≤wu+(1−w)v for every w∈[0,1]w\in[0,1]w∈[0,1]. Moreover, L(0)=−D0(X,Y)L(0)=-D_0(X,Y)L(0)=−D0​(X,Y), L(1)=−D0(Y,X)L(1)=-D_0(Y,X)L(1)=−D0​(Y,X), and, for every a∈[0,1)a\in[0,1)a∈[0,1),

−Da(X,Y)=sup⁡r∈R(−ra1−a−H(r)).-D_a(X,Y)=\sup_{r\in\mathbb R}\left(-\frac{ra}{1-a}-H(r)\right).−Da​(X,Y)=r∈Rsup​(−1−ara​−H(r)).

If tr⁡(SXSY)>0\operatorname{tr}(S_XS_Y)>0tr(SX​SY​)>0, then ψ\psiψ is convex on all of R\mathbb RR, PPP is extended convex in the same finite-upper-bound sense and is lower semicontinuous, its real-valued conversion has derivative

ddsP(s).toReal=ψ ⁣(s1+s)+ψ′ ⁣(s1+s)1+s(s>0),\frac{d}{ds}P(s).\mathrm{toReal} =\psi\!\left(\frac{s}{1+s}\right) +\frac{\psi'\!\left(\frac{s}{1+s}\right)}{1+s}\qquad(s>0),dsd​P(s).toReal=ψ(1+ss​)+1+sψ′(1+ss​)​(s>0),

has derivative within [0,∞)[0,\infty)[0,∞) at 000 equal to ψ(0)+ψ′(0)\psi(0)+\psi'(0)ψ(0)+ψ′(0), and its derivatives converge to ψ(1)\psi(1)ψ(1) as s→+∞s\to+\inftys→+∞; also,

−r<ψ(0)+ψ′(0) ⟹ H(r)=−L(0),ψ(1)<−r ⟹ H(r)=+∞,-r<\psi(0)+\psi'(0)\ \Longrightarrow\ H(r)=-L(0),\qquad \psi(1)<-r\ \Longrightarrow\ H(r)=+\infty,−r<ψ(0)+ψ′(0) ⟹ H(r)=−L(0),ψ(1)<−r ⟹ H(r)=+∞,

P(s)=P∗∗(s)P(s)=P^{**}(s)P(s)=P∗∗(s) for every s≥0s\ge0s≥0, and the displayed supremum for −Da(X,Y)-D_a(X,Y)−Da​(X,Y) can be restricted to

−ψ(1)≤r≤−ψ(0)−ψ′(0).-\psi(1)\le r\le-\psi(0)-\psi'(0).−ψ(1)≤r≤−ψ(0)−ψ′(0).

If tr⁡X≤1\operatorname{tr}X\le1trX≤1, including X=0X=0X=0, then a↦Da(X,Y)a\mapsto D_a(X,Y)a↦Da​(X,Y) is nondecreasing on [0,1)[0,1)[0,1), Da(X,Y)→H(0)D_a(X,Y)\to H(0)Da​(X,Y)→H(0) as a→1−a\to1^-a→1−, H(r)→H(0)H(r)\to H(0)H(r)→H(0) as r→0+r\to0^+r→0+, and H(0)=D1−(X,Y)H(0)=D_{1^-}(X,Y)H(0)=D1−​(X,Y), where D1−D_{1^-}D1−​ is the definition’s totalized left-limit value of DaD_aDa​. Finally, if tr⁡X=1\operatorname{tr}X=1trX=1, then D1−(X,Y)D_{1^-}(X,Y)D1−​(X,Y) equals Re⁡Tr⁡[X(log⁡X−log⁡Y)]\operatorname{Re}\operatorname{Tr}[X(\log X-\log Y)]ReTr[X(logX−logY)] when ker⁡Y⊆ker⁡X\ker Y\subseteq\ker XkerY⊆kerX, and equals +∞+\infty+∞ otherwise, with matrix logarithms interpreted by spectral functional calculus.

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