Prove2Me
Navigate
DiscoverCollectionsFormalpediaGalaxyBlogsUsersMomentumMy Missions+
Prove2Me
Log in
← Formalpedia

Corollary 3.4.2 — Probabilistic Lieb inequality

Proved
TroppMatrixConcentration.ch3_probabilistic_lieb

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

matrix-concentrationrandom-matrices

Let HHH be a fixed Hermitian matrix and XXX a measurable, almost-surely Hermitian random matrix of the same positive dimension on a probability space. If eXe^XeX is Bochner integrable, then

Etr⁡exp⁡(H+X)≤tr⁡exp⁡ ⁣(H+log⁡EeX).\mathbb E\operatorname{tr}\exp(H+X)\le\operatorname{tr}\exp\!\left(H+\log\mathbb E e^X\right).Etrexp(H+X)≤trexp(H+logEeX).

This is the probabilistic form of Lieb's concavity theorem used in the trace-cgf subadditivity argument. No independence hypothesis or integrability of XXX itself is required.

Preamble
import Definitions.Def_TroppMatrixConcentration_probability

open MeasureTheory ProbabilityTheory
open scoped Matrix.Norms.L2Operator

Formal statement
namespace TroppMatrixConcentration

theorem ch3_probabilistic_lieb {Ω : Type*} [MeasurableSpace Ω]
    (μ : Measure Ω) [IsProbabilityMeasure μ] {d : ℕ} [NeZero d]
    (H : Matrix (Fin d) (Fin d) ℂ) (hH : H.IsHermitian)
    (X : Ω → Matrix (Fin d) (Fin d) ℂ)
    (hMeas : Measurable X) (hHerm : ∀ᵐ ω ∂μ, (X ω).IsHermitian)
    (hExp : Integrable (fun ω => matrixExp (X ω)) μ) :
    (∫ ω, traceExp (H + X ω) ∂μ) ≤
      traceExp (H + matrixLog (∫ ω, matrixExp (X ω) ∂μ)) := 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; Corollary 3.4.2, printed p. 35.
Read-back

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

For every measurable space Ω\OmegaΩ, probability measure μ\muμ on Ω\OmegaΩ, natural number d≠0d\ne 0d=0, Hermitian matrix H∈Cd×dH\in\mathbb C^{d\times d}H∈Cd×d, and Borel-measurable matrix-valued function X:Ω→Cd×dX:\Omega\to\mathbb C^{d\times d}X:Ω→Cd×d such that X(ω)X(\omega)X(ω) is Hermitian for μ\muμ-almost every ω\omegaω and ω↦exp⁡(X(ω))\omega\mapsto\exp(X(\omega))ω↦exp(X(ω)) is Bochner integrable with respect to μ\muμ, the following real inequality holds:

∫ΩRe⁡tr⁡(exp⁡(H+X(ω))) dμ(ω)  ≤  Re⁡tr⁡(exp⁡(H+L(∫Ωexp⁡(X(ω)) dμ(ω)))).\int_\Omega \operatorname{Re}\operatorname{tr}\bigl(\exp(H+X(\omega))\bigr)\,d\mu(\omega) \;\le\; \operatorname{Re}\operatorname{tr}\left(\exp\left(H+L\left(\int_\Omega\exp(X(\omega))\,d\mu(\omega)\right)\right)\right).∫Ω​Retr(exp(H+X(ω)))dμ(ω)≤Retr(exp(H+L(∫Ω​exp(X(ω))dμ(ω)))).

Here exp⁡(A)\exp(A)exp(A) denotes the normed-algebra matrix exponential, and L(A)L(A)L(A) is the real continuous functional calculus applied to the scalar function Real.log⁡\operatorname{Real.log}Real.log and the matrix AAA; for Hermitian AAA this applies that real logarithm to its eigenvalues. The scalar logarithm is total, with Real.log⁡(0)=0\operatorname{Real.log}(0)=0Real.log(0)=0 and Real.log⁡(t)=log⁡∣t∣\operatorname{Real.log}(t)=\log|t|Real.log(t)=log∣t∣ for nonzero real ttt; the real continuous functional calculus returns the zero matrix when its applicability conditions fail. All integrals displayed are Bochner integrals. There is no separate hypothesis that the real-valued integrand on the left is integrable, and its integral has the Bochner convention of being zero if it is not integrable. Hermiticity of XXX is required only almost everywhere; no pointwise Hermiticity on exceptional null sets is required. The dimension-zero case is excluded, while d=1d=1d=1 is included; no separate positivity or invertibility hypothesis is imposed on the matrix inside LLL.

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