A measurable family of Gaussian states
OpenBosonicLearn.gaussianState_measurable_familyGaussian states can be chosen measurably in their first and second moments.
For every there is a map from pairs , , , to density operators on the -mode Fock space such that
- is measurable, for the Borel sets of and the Borel sets of the trace-norm topology on density operators;
- every value is a Gaussian state (with some mean and covariance);
- whenever there exists a Gaussian state with mean and covariance matrix , the value is a Gaussian state with mean and covariance .
Why it holds. The pairs that are moments of a Gaussian state are those with symmetric and , a closed set; on it is continuous in trace norm (pointwise convergence of characteristic functions of states to that of a state implies trace-norm convergence). Elsewhere take the vacuum.
Role. The estimator of Theorem 8 outputs , where are measurable functions of the heterodyne outcomes; the statement of Theorem 8 requires the estimator to be a measurable map into states all of whose values are Gaussian states. This lemma supplies it, including on outcomes for which is not a valid covariance matrix.
Formalization Note IsGaussianState ρ μ S is the mission's predicate for : the characteristic function of is , together with finite energy, symmetry of and the first and second moment identities; quadratures are in grouped order and the vacuum has . States are State (Occupation (Fin n)), density operators on . The covariance argument of gs is a function Fin n ⊕ Fin n → Fin n ⊕ Fin n → ℝ, so that the product -algebra of function spaces applies (Matrix carries no measurable structure in Mathlib). The source uses the family without comment.
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 MeasureTheory open scoped ENNReal Matrix
namespace BosonicLearn
/-- **A measurable family of Gaussian states `(µ, Σ) ↦ ρ(µ, Σ)`.**
Chen, Anna Mele, Anna Mele, Preskill, *When are bosonic Gaussian states classical to learn?*, arXiv:2609.26705, §1.1, p. 5: "Such a state is completely determined by its first and second moments … thus we denote a
Gaussian state by ρ(µ, Σ)"; Theorem 8, p. 61: "a non-adaptive protocol … that always outputs a valid Gaussian
state ρ̂", with `ρ̂ = ρ(µ̂, Σ̃)` computed from the measurement outcomes (Eq. (211)).
There is a map `gs` from pairs (vector `µ ∈ ℝ²ⁿ`, array `S ∈ ℝ^{2n×2n}`) to density operators on the `n`-mode Fock
space such that
1. `gs` is measurable (Borel sets of `ℝ²ⁿ × ℝ^{2n×2n}`; trace-norm Borel sets of states);
2. every value of `gs` is a Gaussian state;
3. whenever a Gaussian state with mean `µ` and covariance `Σ` exists, `gs (µ, Σ)` is such a state.
**Formalization Note.** The covariance argument is a function `Fin n ⊕ Fin n → Fin n ⊕ Fin n → ℝ` (grouped
quadratures) so that the product σ-algebra of functions is used. On pairs that are not the moments of a
Gaussian state the value is some Gaussian state (for instance the vacuum). Measurability follows from
continuity of `(µ, Σ) ↦ ρ(µ, Σ)` in trace norm on the closed set of valid pairs. -/
theorem gaussianState_measurable_family (n : ℕ) :
∃ gs : (Fin n ⊕ Fin n → ℝ) × (Fin n ⊕ Fin n → Fin n ⊕ Fin n → ℝ) → State (Occupation (Fin n)),
Measurable gs ∧
(∀ p, ∃ (μ : Fin n ⊕ Fin n → ℝ) (S : Matrix (Fin n ⊕ Fin n) (Fin n ⊕ Fin n) ℝ),
IsGaussianState (gs p) μ S) ∧
∀ (μ : Fin n ⊕ Fin n → ℝ) (S : Matrix (Fin n ⊕ Fin n) (Fin n ⊕ Fin n) ℝ)
(ρ : State (Occupation (Fin n))), IsGaussianState ρ μ S →
IsGaussianState (gs (μ, fun a b => S a b)) μ S := by sorry
end BosonicLearn