Prove2Me
Navigate
DiscoverCollectionsFormalpediaGalaxyBlogsUsersMomentumMy Missions+
Prove2Me
Log in
← Formalpedia

Theorem 8.6.2 — Matrix perspective is jointly operator convex

Proved
TroppMatrixConcentration.ch8_perspective_convex

by tc · Oct 7, 2026 · Mathlib 0df444a (Lean v4.33.1)

convexitymatrix-analysismatrix-concentration

Let fff be operator convex on (0,∞)(0,\infty)(0,∞) and let A1,A2,H1,H2A_1,A_2,H_1,H_2A1​,A2​,H1​,H2​ be positive-definite complex matrices of the same positive dimension. For every 0≤t≤10\le t\le10≤t≤1,

Ψf(tA1+(1−t)A2; tH1+(1−t)H2)⪯tΨf(A1;H1)+(1−t)Ψf(A2;H2).\Psi_f(tA_1+(1-t)A_2;\,tH_1+(1-t)H_2)\preceq t\Psi_f(A_1;H_1)+(1-t)\Psi_f(A_2;H_2).Ψf​(tA1​+(1−t)A2​;tH1​+(1−t)H2​)⪯tΨf​(A1​;H1​)+(1−t)Ψf​(A2​;H2​).

Here Ψf(A;H)=A1/2f(A−1/2HA−1/2)A1/2\Psi_f(A;H)=A^{1/2}f(A^{-1/2}HA^{-1/2})A^{1/2}Ψf​(A;H)=A1/2f(A−1/2HA−1/2)A1/2. No pair of the matrices is required to commute. This supplies the bivariate operator-convexity result used in the entropy analysis.

Preamble
import Definitions.Def_TroppMatrixConcentration_ch8_entropy

open scoped Matrix.Norms.L2Operator ComplexOrder

Formal statement
namespace TroppMatrixConcentration

theorem ch8_perspective_convex {d : ℕ} [NeZero d]
    (f : ℝ → ℝ) (hf : ch8_operatorConvexOn (Set.Ioi 0) f)
    (A₁ A₂ H₁ H₂ : Matrix (Fin d) (Fin d) ℂ)
    (hA₁ : A₁.PosDef) (hA₂ : A₂.PosDef) (hH₁ : H₁.PosDef) (hH₂ : H₂.PosDef)
    (t : ℝ) (ht₀ : 0 ≤ t) (ht₁ : t ≤ 1) :
    loewnerLE
      (ch8_perspective f (t • A₁ + (1 - t) • A₂) (t • H₁ + (1 - t) • H₂))
      (t • ch8_perspective f A₁ H₁ + (1 - t) • ch8_perspective f A₂ H₂) := by sorry

end TroppMatrixConcentration
Source
Joel A. Tropp, An Introduction to Matrix Concentration Inequalities, arXiv:1501.01571v1 (7 January 2015); https://arxiv.org/abs/1501.01571v1; Theorem 8.6.2, printed pp. 134–135.
Read-back

What the Lean code literally says, in plain math · gpt-6

For every natural number d≠0d\ne 0d=0, every function f:R→Rf:\mathbb R\to\mathbb Rf:R→R, every four positive definite complex d×dd\times dd×d matrices A1,A2,H1,H2A_1,A_2,H_1,H_2A1​,A2​,H1​,H2​, and every real ttt with 0≤t≤10\le t\le 10≤t≤1, the following implication holds. Write Fg(X)F_g(X)Fg​(X) for the real continuous functional calculus cfc⁡(g,X)\operatorname{cfc}(g,X)cfc(g,X), using its total-function convention (zero when the required self-adjointness or continuity on the real spectrum fails), and write X⪯YX\preceq YX⪯Y to mean that Y−XY-XY−X is positive semidefinite. Assume that (0,∞)(0,\infty)(0,∞) is convex and that, for every nonzero natural number nnn, every pair of Hermitian complex n×nn\times nn×n matrices B,CB,CB,C whose real spectra are contained in (0,∞)(0,\infty)(0,∞), and every real sss with 0≤s≤10\le s\le10≤s≤1, one has Ff(sB+(1−s)C)⪯sFf(B)+(1−s)Ff(C)F_f(sB+(1-s)C)\preceq sF_f(B)+(1-s)F_f(C)Ff​(sB+(1−s)C)⪯sFf​(B)+(1−s)Ff​(C). Define SX=Fx↦x(X)S_X=F_{x\mapsto\sqrt{x}}(X)SX​=Fx↦x​​(X), RX=Fx↦(x)−1(X)R_X=F_{x\mapsto(\sqrt{x})^{-1}}(X)RX​=Fx↦(x​)−1​(X), and Pf(X,Y)=SXFf(RXYRX)SXP_f(X,Y)=S_XF_f(R_XYR_X)S_XPf​(X,Y)=SX​Ff​(RX​YRX​)SX​, with the displayed products interpreted as matrix multiplication. Then Pf(tA1+(1−t)A2, tH1+(1−t)H2)⪯tPf(A1,H1)+(1−t)Pf(A2,H2)P_f(tA_1+(1-t)A_2,\,tH_1+(1-t)H_2)\preceq tP_f(A_1,H_1)+(1-t)P_f(A_2,H_2)Pf​(tA1​+(1−t)A2​,tH1​+(1−t)H2​)⪯tPf​(A1​,H1​)+(1−t)Pf​(A2​,H2​). Thus the difference between the right-hand and left-hand matrices is positive semidefinite. Positive definiteness here includes Hermitian symmetry and strict positivity of the associated quadratic form on every nonzero complex vector; positive semidefiniteness includes Hermitian symmetry and nonnegative quadratic form on every vector. All scalar coefficients are real. The matrix-size quantifiers exclude size zero, the weight intervals include both endpoints, and there is no separately stated continuity assumption on fff.

Human review
  • Endorsed by Shuze Chen · Oct 7, 2026

    Confirmed by the moderator at approval.

  • Endorsed by tc · Oct 7, 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