Trace-norm separation of passive Gaussian states from their occupation matrices, uniform in temperature:
ProvedBosonicLearn.occupation_count_separationThe pairwise-distance bound for the hard ensemble of Theorem 9, by the photon-counting argument of Lemma 12, with a constant independent of the temperature scale .
There is an absolute constant such that: for , with , and Hermitian occupation matrices with , and , the passive Gaussian states with fugacity matrices satisfy
With , and the right-hand side is .
Proof idea: count photons in the positive eigen-directions of ; the two counts are sums of geometric variables with paired means (BosonicLearn.loewner_eigenvalue_pairing, BosonicLearn.fugacity_count_pgf), a threshold separates them (BosonicLearn.geometric_threshold_scaled), and data processing (BosonicLearn.fugacity_count_test) transfers this to the trace norm.
Formalization Note is the full trace norm traceSize; the lower bound is ENNReal.ofReal (…).
import Mathlib import Definitions.Def_BosonicLearn_fugacityCoeff import Definitions.Def_QAlgorithms_Papers_Independent_Dependency50_P33 open QAlgorithms.Papers.Independent.Dependency50.P33 (traceSize) open QAlgorithms.Papers.Independent.Claimed50.P47 (Occupation) open scoped ENNReal Matrix ComplexOrder
namespace BosonicLearn
/-- **Trace-norm separation of passive Gaussian states from their occupation matrices, uniformly in temperature.**
Chen, Anna Mele, Anna Mele, Preskill, *When are bosonic Gaussian states classical to learn?*, arXiv:2609.26705, §8, proof of Theorem 9, Eqs. (226)–(228), p. 66 (pairwise trace distance of the hard ensemble for `ν ≥ 1`), by the
photon-counting argument of §5.7 (proof of Lemma 12, Eqs. (135)–(148)).
There is an absolute constant `c > 0` with the following property. Let `m ≥ 1` and let `N₁, N₂` be Hermitian `n × n`
occupation matrices with `(7/8)m·1 ⪯ N_i ⪯ (9/8)m·1`, `Tr(N₁ − N₂) = 0` and `−θm·1 ⪯ N₁ − N₂ ⪯ θm·1`, where
`θ√n ≤ 7/8`. Then the passive Gaussian states `ρ_i = det(1 − R_i)Γ(R_i)` with fugacity matrices
`R_i = N_i(1 + N_i)⁻¹` satisfy
`‖ρ₁ − ρ₂‖₁ ≥ c · Tr (N₁ − N₂)² / (θ m² √n)`.
**Formalization Note.** `‖·‖₁` is the full trace norm `traceSize`. With `N_i = m(1 + ξX_i)`, `‖X_i‖_op ≤ 4` and
`θ = 8ξ` the right-hand side is `(c/8) ξ ‖X₁ − X₂‖_F² / √n`; the bound does not depend on `m`. -/
theorem occupation_count_separation :
∃ c : ℝ, 0 < c ∧ ∀ (n : ℕ) (m θ : ℝ), 1 ≤ m → 0 < θ → θ * Real.sqrt n ≤ 7 / 8 →
∀ N₁ N₂ : Matrix (Fin n) (Fin n) ℂ,
(N₁ - ((7 / 8 * m : ℝ) : ℂ) • (1 : Matrix (Fin n) (Fin n) ℂ)).PosSemidef →
(((9 / 8 * m : ℝ) : ℂ) • (1 : Matrix (Fin n) (Fin n) ℂ) - N₁).PosSemidef →
(N₂ - ((7 / 8 * m : ℝ) : ℂ) • (1 : Matrix (Fin n) (Fin n) ℂ)).PosSemidef →
(((9 / 8 * m : ℝ) : ℂ) • (1 : Matrix (Fin n) (Fin n) ℂ) - N₂).PosSemidef →
(N₁ - N₂).trace = 0 →
(((θ * m : ℝ) : ℂ) • (1 : Matrix (Fin n) (Fin n) ℂ) - (N₁ - N₂)).PosSemidef →
(((θ * m : ℝ) : ℂ) • (1 : Matrix (Fin n) (Fin n) ℂ) + (N₁ - N₂)).PosSemidef →
ENNReal.ofReal (c * ((N₁ - N₂) * (N₁ - N₂)).trace.re / (θ * m ^ 2 * Real.sqrt n)) ≤
traceSize (fugacityCoeff (N₁ * (1 + N₁)⁻¹) - fugacityCoeff (N₂ * (1 + N₂)⁻¹)) := by sorry
end BosonicLearn