hermitian_kth_eigenvalue_witness
Provedsensitivity_theorem
Min-max witness (>=) for the k-th descending eigenvalue: every subspace W of (V -> R) of dimension at least (card V) - k contains a nonzero v with Rayleigh quotient >= lambda_k. Generalizes Aristole's exists_eigenvector_in_subspace from the specific eigenvalue +sqrt(n).
Preamble
import Mathlib.Analysis.Matrix.Spectrum import Mathlib.LinearAlgebra.Matrix.Hermitian import Mathlib.LinearAlgebra.FiniteDimensional.Defs import Mathlib.Data.Matrix.Mul import Mathlib.Data.Fintype.Basic import Mathlib.Data.Real.Star open Matrix /-! # Generalized min-max characterization (lower-bound witness form) For any Hermitian real matrix `A : Matrix V V ℝ`, descending-sorted index `k`, and subspace `W ⊆ (V → ℝ)` whose dimension is at least `(card V) − k`, there exists a nonzero vector `v ∈ W` whose Rayleigh quotient is at least the `k`-th descending eigenvalue of `A`. This is the existential, witness-form *upper-half* min-max characterization (one direction of Courant–Fischer). It generalizes Aristole's `exists_eigenvector_in_subspace`, which is the same statement at the specific eigenvalue `+√n`. Reusable spectral-theory result, not currently in Mathlib in this form. -/ /-- **Min-max witness for the `k`-th descending eigenvalue.** For any real Hermitian `A : Matrix V V ℝ` and any subspace `W ⊆ (V → ℝ)` with `Module.finrank ℝ W ≥ (card V) − k`, there exists a nonzero `v ∈ W` whose Rayleigh quotient `(A *ᵥ v ⬝ᵥ v) / (v ⬝ᵥ v)` is at least `hA.eigenvalues₀ k`. -/
Formal statement
theorem hermitian_kth_eigenvalue_witness
{V : Type*} [Fintype V] [DecidableEq V] [Nonempty V]
{A : Matrix V V ℝ} (hA : A.IsHermitian) (k : Fin (Fintype.card V))
{W : Submodule ℝ (V → ℝ)}
(hdim : Fintype.card V - (k : ℕ) ≤ Module.finrank ℝ W) :
∃ v : V → ℝ, v ∈ W ∧ v ≠ 0 ∧
hA.eigenvalues₀ k * (v ⬝ᵥ v) ≤ A *ᵥ v ⬝ᵥ v := by sorrySource
Huang, Hao. "Induced subgraphs of hypercubes and a proof of the Sensitivity Conjecture." Annals of Mathematics 190.3 (2019): 949-955.