Theorem 3.1 — Schur-complement sufficient condition and P-matrix characterization
ProvedRobinsonSR.Schur.theorem_3_1Let be positive integers, let be nonempty, closed, and convex, and partition into , , , and blocks. Define . If is nonsingular and its Schur complement is positive definite in the sense for nonzero , then is an everywhere-defined, single-valued, globally Lipschitz function. In the special case , this has an exact characterization:
The first assertion covers arbitrary nonempty closed convex ; the second replaces positive definiteness by the P-matrix condition on the orthant and asserts its necessity. Together these are Robinson's Theorem 3.1.
Formalization Note The Schur complement is only used under the nonsingularity condition; positive definiteness does not require a symmetric matrix. The Lean statement quantifies over all eligible for the first assertion and uses the orthant for the second.
import Mathlib import Definitions.Def_RobinsonSR_Schur_Setting
namespace RobinsonSR.Schur
/-- Robinson, Theorem 3.1, p. 51. -/
theorem theorem_3_1 (r s : ℕ) (hr : 0 < r) (hs : 0 < s)
(A₁₁ : Matrix (Fin r) (Fin r) ℝ)
(A₁₂ : Matrix (Fin r) (Fin s) ℝ)
(A₂₁ : Matrix (Fin s) (Fin r) ℝ)
(A₂₂ : Matrix (Fin s) (Fin s) ℝ) :
(∀ K : Set (EuclideanSpace ℝ (Fin s)), K.Nonempty → IsClosed K → Convex ℝ K →
A₁₁.det ≠ 0 → PosDefNS (schur A₁₁ A₁₂ A₂₁ A₂₂) →
InvIsLipschitzFunction (Matrix.fromBlocks A₁₁ A₁₂ A₂₁ A₂₂) (prodSet K)) ∧
(InvIsLipschitzFunction (Matrix.fromBlocks A₁₁ A₁₂ A₂₁ A₂₂)
(prodSet (nonnegOrthant s)) ↔
A₁₁.det ≠ 0 ∧ IsPMatrix (schur A₁₁ A₁₂ A₂₁ A₂₂)) := by sorry
end RobinsonSR.Schur
Confirmed by the mission captain (proposal self-audit).
Confirmed by the moderator at approval.