for positive definite
ProvedBosonicLearn.loewner_inv_form_lebosonic-gaussian-statesquantum-informationquantum-learning
Inversion reverses the Loewner order, in quadratic-form version.
Let be real positive definite matrices of the same size and with . Then for every real vector ,
Role. In the proof of Theorem 8, is the covariance of the heterodyne outcomes, and : the concentration of in the norm (Eq. (210)) gives the bound in the norm needed by Lemma 18.
Formalization Note The quadratic forms are v ⬝ᵥ (B⁻¹ *ᵥ v); Loewner order is Matrix.PosSemidef of the difference. Pure linear algebra.
Preamble
import Mathlib open scoped Matrix
Formal statement
namespace BosonicLearn
/-- **Inversion reverses the Loewner order (quadratic-form version).**
Chen, Anna Mele, Anna Mele, Preskill, *When are bosonic Gaussian states classical to learn?*, arXiv:2609.26705, §7.2, proof of Theorem 8, p. 62: "From C ⪯ α_ν Σ, inversion reverses the positive-semidefinite order, so
C^{−1} ⪰ α_ν^{−1} Σ^{−1}, or equivalently Σ^{−1} ⪯ α_ν C^{−1}. Consequently,
‖Σ^{−1/2}(µ̂ − µ)‖₂ ≤ √α_ν ‖C^{−1/2}(µ̂ − µ)‖₂."
Let `A, B` be real positive definite matrices with `A ⪯ α B`, `α > 0`. Then for every vector `v`,
`vᵀ B⁻¹ v ≤ α · vᵀ A⁻¹ v`.
**Formalization Note.** `‖M^{−1/2} v‖₂² = vᵀ M⁻¹ v` for positive definite `M`; Loewner order is positive
semidefiniteness of the difference. -/
theorem loewner_inv_form_le {ι : Type} [Fintype ι] [DecidableEq ι] (A B : Matrix ι ι ℝ) (hA : A.PosDef)
(hB : B.PosDef) (α : ℝ) (hα : 0 < α) (hAB : (α • B - A).PosSemidef) (v : ι → ℝ) :
v ⬝ᵥ (B⁻¹ *ᵥ v) ≤ α * (v ⬝ᵥ (A⁻¹ *ᵥ v)) := by sorry
end BosonicLearn
Source
Chen, Anna Mele, Anna Mele, Preskill, When are bosonic Gaussian states classical to learn?, arXiv:2609.26705, p. 62, between Eqs. (216) and (217) ("inversion reverses the positive-semidefinite order", proof of Theorem 8)