Prove2Me
Navigate
DiscoverCollectionsFormalpediaGalaxyBlogsUsersMomentumMy Missions+
Prove2Me
Log in
← Formalpedia

Theorem 8.5.2 — Operator Jensen inequality

Proved
TroppMatrixConcentration.ch8_operator_jensen

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

convexitymatrix-analysismatrix-concentration

Let fff be operator convex on a real interval III. Let A1,A2A_1,A_2A1​,A2​ be Hermitian complex matrices with spectra in III, and let K1,K2K_1,K_2K1​,K2​ be compatible rectangular complex matrices with the same positive number of columns and

K1∗K1+K2∗K2=I.K_1^*K_1+K_2^*K_2=I.K1∗​K1​+K2∗​K2​=I.

Then

f(K1∗A1K1+K2∗A2K2)⪯K1∗f(A1)K1+K2∗f(A2)K2.f(K_1^*A_1K_1+K_2^*A_2K_2)\preceq K_1^*f(A_1)K_1+K_2^*f(A_2)K_2.f(K1∗​A1​K1​+K2∗​A2​K2​)⪯K1∗​f(A1​)K1​+K2∗​f(A2​)K2​.

All matrix dimensions are positive. This extends scalar convex combinations to matrix convex combinations; the two input matrices may have different compatible sizes.

Preamble
import Definitions.Def_TroppMatrixConcentration_ch8_entropy

open scoped Matrix.Norms.L2Operator ComplexOrder

Formal statement
namespace TroppMatrixConcentration

theorem ch8_operator_jensen {d m n : ℕ} [NeZero d] [NeZero m] [NeZero n]
    (I : Set ℝ) (f : ℝ → ℝ) (hf : ch8_operatorConvexOn I f)
    (A₁ : Matrix (Fin m) (Fin m) ℂ) (A₂ : Matrix (Fin n) (Fin n) ℂ)
    (hA₁ : A₁.IsHermitian) (hA₂ : A₂.IsHermitian)
    (hA₁I : spectrum ℝ A₁ ⊆ I) (hA₂I : spectrum ℝ A₂ ⊆ I)
    (K₁ : Matrix (Fin m) (Fin d) ℂ) (K₂ : Matrix (Fin n) (Fin d) ℂ)
    (hK : K₁.conjTranspose * K₁ + K₂.conjTranspose * K₂ = 1) :
    loewnerLE
      (ch8_matrixFunction f (K₁.conjTranspose * A₁ * K₁ + K₂.conjTranspose * A₂ * K₂))
      (K₁.conjTranspose * ch8_matrixFunction f A₁ * K₁ +
        K₂.conjTranspose * ch8_matrixFunction f A₂ * K₂) := 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 8.5.2 and Chapter 8 dimension conventions, printed pp. 120, 132–133.
Read-back

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

For every three positive integers d,m,nd,m,nd,m,n, every set I⊆RI\subseteq\mathbb RI⊆R, and every function f:R→Rf:\mathbb R\to\mathbb Rf:R→R, suppose that III is convex and that the following property holds in every positive integer dimension rrr: for all Hermitian complex r×rr\times rr×r matrices A,HA,HA,H whose real spectra are contained in III, and every real ttt with 0≤t≤10\le t\le10≤t≤1, the matrix tf(A)+(1−t)f(H)−f(tA+(1−t)H)t f(A)+(1-t)f(H)-f(tA+(1-t)H)tf(A)+(1−t)f(H)−f(tA+(1−t)H) is positive semidefinite. Here and throughout, f(B)f(B)f(B) means the real continuous functional calculus applied to fff and BBB; for a finite Hermitian matrix this applies fff to its eigenvalues. Let A1A_1A1​ and A2A_2A2​ be Hermitian complex matrices of sizes m×mm\times mm×m and n×nn\times nn×n, respectively, with both real spectra contained in III. For every complex m×dm\times dm×d matrix K1K_1K1​ and complex n×dn\times dn×d matrix K2K_2K2​ satisfying K1∗K1+K2∗K2=IdK_1^*K_1+K_2^*K_2=I_dK1∗​K1​+K2∗​K2​=Id​, where ∗^*∗ denotes conjugate transpose and IdI_dId​ is the identity matrix, the matrix

K1∗f(A1)K1+K2∗f(A2)K2−f(K1∗A1K1+K2∗A2K2)K_1^*f(A_1)K_1+K_2^*f(A_2)K_2-f\bigl(K_1^*A_1K_1+K_2^*A_2K_2\bigr)K1∗​f(A1​)K1​+K2∗​f(A2​)K2​−f(K1∗​A1​K1​+K2∗​A2​K2​)

is positive semidefinite, meaning that it is Hermitian and its quadratic form is nonnegative on every complex vector in dimension ddd. All three matrix dimensions, and every dimension quantified in the hypothesis on fff, exclude zero; the weight hypothesis includes t=0t=0t=0 and t=1t=1t=1. No separate continuity assumption on fff, nor openness, closedness, or nonemptiness assumption on III, is imposed. If III is empty, the required spectral containment for the positive dimensional Hermitian input matrices cannot hold.

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