Sylvester's inequality (hard direction): for subspaces where is positive definite
ProvedRHLinalg.finrank_le_posIndex_of_posDefOnLet be an Hermitian matrix over ( or , an RCLike field), and let (posIndex) denote its positive index: the number of strictly positive eigenvalues. Let be a subspace on which the Hermitian form is positive definite, i.e. for every nonzero (PosDefOn).
Statement.
This is the hard direction of the subspace characterization of the positive index from Sylvester's law of inertia: no subspace on which the form is positive definite can exceed the number of positive eigenvalues. Together with the easy direction RHLinalg.posDefOn_range_hermPosPart (the form is positive definite on a subspace of dimension exactly ), it makes the inertia lemma of the paper (docstring reference lem:inertia) immediate. In the module Zeta23.LinAlg.Sylvester it is consumed by Zeta23.ZeroSide.posIndex_smul_pos in the zero-side positivity analysis of the matrix-variational argument.
import Mathlib.Analysis.Matrix.PosDef
import Mathlib.LinearAlgebra.FiniteDimensional.Lemmas
import Definitions.Def_Zeta23_LinAlg_HermitianPosPart
import Definitions.Def_Zeta23_LinAlg_PosIndex
import Definitions.Def_Zeta23_LinAlg_Sylvester
open Matrix Finset Submodule
open scoped ComplexOrder
open RHLinalg
variable {𝕜 : Type*} [RCLike 𝕜]
variable {n : Type*} [Fintype n] [DecidableEq n]
theorem RHLinalg.finrank_le_posIndex_of_posDefOn {A : Matrix n n 𝕜} (hA : A.IsHermitian)
{W : Submodule 𝕜 (n → 𝕜)} (hW : PosDefOn A W) :
Module.finrank 𝕜 W ≤ posIndex hA := by sorry