[BMEM25, Theorem 8], covariance part: for
OpenBosonicLearn.bmem25_theorem_8_covA trace-distance bound for two Gaussian states with the same mean and ordered covariance matrices, without any gap above the vacuum.
Let and be -mode Gaussian states with the same mean and . Then
Source. This is the covariance part of Theorem 8 of Bittel, Mele, Eisert, Mele, Energy-independent tomography of Gaussian states, arXiv:2508.14979 ([BMEM25] in the paper): , specialized to equal means and . The right-hand side is invariant under a common rescaling of the two covariance matrices, so it reads the same in the convention of [BMEM25] (vacuum ) and in the convention of this mission (vacuum ).
Role. It is the trace-distance input of the gap-free heterodyne bound BosonicLearn.bmem25_theorem_10 ( copies), used for in Corollary 19: with the right-hand side is at most , linear in , where Theorem 7 of the paper gives under a vacuum gap.
Formalization Note The states are given through IsGaussianState; the hypothesis is Matrix.PosSemidef (S' − S); the bound is ENNReal.ofReal of the right-hand side with Matrix.trace and matrix inverses (covariance matrices of states are positive definite). The mean term of the cited theorem is not included; in the application it is handled by Lemma 18 of the paper.
import Mathlib import Definitions.Def_BosonicLearn_traceDist import Definitions.Def_BosonicLearn_IsGaussianState open QAlgorithms.Papers.Independent.Dependency50.P33 (State traceSize) open QAlgorithms.Papers.Independent.Claimed50.P47 (Occupation) open scoped ENNReal Matrix
namespace BosonicLearn
/-- **[BMEM25, Theorem 8], covariance part, for ordered covariances.**
Chen, Anna Mele, Anna Mele, Preskill, *When are bosonic Gaussian states classical to learn?*, arXiv:2609.26705, §1.4, p. 10: "an important foundation for these results is the recently developed set of bounds on the
trace distance between Gaussian states [BMM+25, Hol24, BMEM25, FRSF26]", and proof of Corollary 19, p. 65: "the
general heterodyne tomography bound of [BMEM25, Theorem 10] applies without using the vacuum-gap promise".
[BMEM25] = Bittel, Mele, Eisert, Mele, *Energy-independent tomography of Gaussian states*, arXiv:2508.14979,
Theorem 8: `‖ρ(V, t) − ρ(W, m)‖₁ ≤ 2‖V^{−1/2}(m − t)‖₂ + ((1 + √3)/4) Tr((V⁻¹ + W⁻¹)|V − W|)`.
For Gaussian states `ρ(µ, Σ′)` and `ρ(µ, Σ)` with the same mean and `Σ′ ⪰ Σ`,
`½‖ρ(µ, Σ′) − ρ(µ, Σ)‖₁ ≤ ((1 + √3)/8) · Tr((Σ⁻¹ + Σ′⁻¹)(Σ′ − Σ))`.
**Formalization Note.** Special case of the cited bound with equal means and `|Σ′ − Σ| = Σ′ − Σ`. The right-hand
side is invariant under a common rescaling of the covariances, so it is the same in the convention `V = 2Σ` of
[BMEM25] (vacuum `V = I`) and in the convention of this paper (vacuum `Σ = ½I`). Covariance matrices of
states are positive definite, so the inverses are the true inverses. -/
theorem bmem25_theorem_8_cov {n : ℕ} (ρ' ρ : State (Occupation (Fin n))) (μ : Fin n ⊕ Fin n → ℝ)
(S' S : Matrix (Fin n ⊕ Fin n) (Fin n ⊕ Fin n) ℝ)
(h' : IsGaussianState ρ' μ S') (h : IsGaussianState ρ μ S) (hord : (S' - S).PosSemidef) :
traceDist ρ'.coefficients ρ.coefficients ≤
ENNReal.ofReal ((1 + Real.sqrt 3) / 8 * ((S⁻¹ + S'⁻¹) * (S' - S)).trace) := by sorry
end BosonicLearn