Fourier transform of the Husimi function: for every state
ProvedBosonicLearn.husimi_fourierbosonic-gaussian-statesquantum-informationquantum-learning
The Husimi function of a state and its Weyl characteristic function.
Let be a state (density operator) on the -mode Fock space, its Husimi function in heterodyne coordinates , and its characteristic function. Then:
- the iterated series defining converge, for every ;
- , and is measurable and integrable on ;
- for every ,
Role. At : the heterodyne outcome law has total mass one. For a Gaussian state the right-hand side is , which is Eq. (207) of the paper (BosonicLearn.eq_207).
Formalization Note is husimiQ ρ.coefficients, is the published Support50.P14.oscillatorCharacteristic at the interleaved copy of , and the convergence statement is OscillatorCharacteristicSummable. The statement holds for every State, not only Gaussian ones.
Preamble
import Mathlib import Definitions.Def_BosonicLearn_husimiQ import Definitions.Def_QAlgorithms_Papers_Independent_Support50_P14 open QAlgorithms.Papers.Independent.Dependency50.P33 (State) open QAlgorithms.Papers.Independent.Claimed50.P47 (Occupation) open QAlgorithms.Papers.Independent.Support50.P14 (oscillatorCharacteristic OscillatorCharacteristicSummable) open MeasureTheory
Formal statement
namespace BosonicLearn
/-- **Fourier transform of the Husimi function: `∫ e^{it·β} Q_ρ(β) dβ = e^{−|t|²/4} Tr(ρ e^{it·r̂})`.**
Chen, Anna Mele, Anna Mele, Preskill, *When are bosonic Gaussian states classical to learn?*, arXiv:2609.26705, §1.1, p. 5: "heterodyne measurement, whose outcome distribution on an n-mode Gaussian state ρ(µ, Σ) is a 2n-variate classical
Gaussian N(µ, Σ + ½I)", and Eq. (207), p. 61.
For every state `ρ` on the `n`-mode Fock space: the Weyl characteristic function `χ_ρ(ξ) = Tr(ρ e^{iξ·r̂})` is
given by convergent iterated series; the Husimi function `Q_ρ(β) = (2π)^{−n}⟨α(β)|ρ|α(β)⟩` is non-negative, measurable
and integrable on `ℝ²ⁿ`; and for all `t ∈ ℝ²ⁿ`
`∫ e^{it·β} Q_ρ(β) dβ = e^{−|t|²/4} χ_ρ(t)`.
**Formalization Note.** `Q_ρ` is `husimiQ ρ.coefficients`, `χ_ρ` is the published
`Support50.P14.oscillatorCharacteristic` at the interleaved copy of `t` (`(j, 0) ↦ t_{q,j}`, `(j, 1) ↦ t_{p,j}`), and the
convergence statement is `OscillatorCharacteristicSummable`. At `t = 0` the identity says that the heterodyne
outcome law has total mass one. -/
theorem husimi_fourier {n : ℕ} (ρ : State (Occupation (Fin n))) :
OscillatorCharacteristicSummable ρ.coefficients ∧
(∀ β : Fin n ⊕ Fin n → ℝ, 0 ≤ husimiQ ρ.coefficients β) ∧
Measurable (husimiQ ρ.coefficients) ∧
Integrable (husimiQ ρ.coefficients) ∧
∀ t : Fin n ⊕ Fin n → ℝ,
∫ β : Fin n ⊕ Fin n → ℝ, Complex.exp (Complex.I * ((∑ a, t a * β a : ℝ) : ℂ)) *
((husimiQ ρ.coefficients β : ℝ) : ℂ) =
((Real.exp (-(∑ a, t a ^ 2) / 4) : ℝ) : ℂ) *
oscillatorCharacteristic ρ.coefficients
(fun a : Fin n × Fin 2 => if a.2 = 0 then t (Sum.inl a.1) else t (Sum.inr a.1)) := by sorry
end BosonicLearn
Source
Chen, Anna Mele, Anna Mele, Preskill, When are bosonic Gaussian states classical to learn?, arXiv:2609.26705, §1.1, p. 5 (heterodyne measurement: the outcome distribution on ρ(µ, Σ) is N(µ, Σ + ½I)); proof of Theorem 8, Eq. (207), p. 61