Gaussian mixing of displaced states: a state with for
ProvedBosonicLearn.state_gaussian_mixturebosonic-gaussian-statesquantum-informationquantum-learning
Additive classical Gaussian noise on a state.
Let be a state on the -mode Fock space, and a positive semidefinite real matrix. Then there is a state — the mixture of displaced copies of — with
Role. Applied to a Gaussian state with , , it produces the Gaussian state (BosonicLearn.gaussianState_exists_of_le).
Formalization Note are in the interleaved quadrature coordinates of Support50.P14.oscillatorCharacteristic; is meanForm m ξ and is varianceForm T ξ.
Preamble
import Mathlib import Definitions.Def_QAlgorithms_Papers_Independent_Support50_P14 open QAlgorithms.Papers.Independent.Dependency50.P33 (State traceSize) open QAlgorithms.Papers.Independent.Claimed50.P47 (Occupation) open QAlgorithms.Papers.Independent.Support50.P14 open scoped ENNReal Matrix
Formal statement
namespace BosonicLearn
/-- **Classical Gaussian mixing of displaced states (additive Gaussian noise).**
Chen, Anna Mele, Anna Mele, Preskill, *When are bosonic Gaussian states classical to learn?*, arXiv:2609.26705, proof of Theorem 8, p. 62 (the estimator is a Gaussian state whose covariance dominates the true one); §1.1, pp. 4–5.
For a state `ρ` on the `n`-mode Fock space, a vector `m ∈ ℝ²ⁿ` and a positive semidefinite real `2n × 2n` matrix
`T`, there is a state `ρ'` (the mixture `∫ D_x ρ D_x† dN(m, T)(x)` of displaced copies of `ρ`) with characteristic
function `χ_{ρ'}(ξ) = exp(i ξ·m − ½ ξᵀTξ) χ_ρ(ξ)`.
**Formalization Note.** `m, ξ, T` are in the interleaved quadrature coordinates of
`Support50.P14.oscillatorCharacteristic`; `ξ·m` is `meanForm m ξ` and `ξᵀTξ` is `varianceForm T ξ`. -/
theorem state_gaussian_mixture {n : ℕ} (ρ : State (Occupation (Fin n))) (m : Fin n × Fin 2 → ℝ)
(T : Matrix (Fin n × Fin 2) (Fin n × Fin 2) ℝ) (hT : T.PosSemidef) :
∃ ρ' : State (Occupation (Fin n)), ∀ ξ : Fin n × Fin 2 → ℝ,
oscillatorCharacteristic ρ'.coefficients ξ =
Complex.exp (Complex.I * (meanForm m ξ : ℂ) - ((varianceForm T ξ / 2 : ℝ) : ℂ)) *
oscillatorCharacteristic ρ.coefficients ξ := by sorry
end BosonicLearn
Source
Chen, Anna Mele, Anna Mele, Preskill, When are bosonic Gaussian states classical to learn?, arXiv:2609.26705, proof of Theorem 8, p. 62 ("a common displacement is implemented by the same unitary on both states, so it does not change their trace distance"; the estimator's covariance dominates the true one)