Passive unitaries rotate displacements: on Fock matrix elements
ProvedBosonicLearn.gammaFock_weylbosonic-gaussian-statesquantum-informationquantum-learning
Covariance of the displacement (Weyl) operators under passive Gaussian unitaries.
Let be a unitary matrix, , and the Fock matrix elements of the -mode displacement operator. For all occupation vectors ,
Role. is the passive Gaussian unitary acting on the mode operators by . With this identity the characteristic function of reduces, by diagonalizing , to a product of single-mode thermal characteristic functions (BosonicLearn.fugacity_characteristic).
Formalization Note is BosonicLearn.gammaFock and is Dependency50.P39.displacementMatrix. preserves the total photon number, so both sums are finite; they are written over the box filtered by , as in BosonicLearn.gammaFock_mul. Unitarity is the hypothesis Uᴴ * U = 1.
Preamble
import Mathlib import Definitions.Def_BosonicLearn_gammaFock import Definitions.Def_QAlgorithms_Papers_Independent_Dependency50_P39 open QAlgorithms.Papers.Independent.Claimed50.P47 (Occupation) open QAlgorithms.Papers.Independent.Dependency50.P39 (displacementMatrix) open scoped Matrix
Formal statement
namespace BosonicLearn
/-- **Passive unitaries rotate displacements: `Γ(U)† D(z) Γ(U) = D(U†z)`.**
Chen, Anna Mele, Anna Mele, Preskill, *When are bosonic Gaussian states classical to learn?*, arXiv:2609.26705, §1.1, pp. 4–5 (displacement operators; passive Gaussian unitaries act on the mode operators by a unitary matrix); Lemma 2,
p. 20 (passive Gaussian states `det(I − R)Γ(R)`).
For a unitary `n × n` matrix `U`, amplitudes `z ∈ ℂⁿ` and occupation vectors `u, v`, with
`D(z)_{ww'} = ∏ⱼ D(zⱼ)_{wⱼ w'ⱼ}`,
`∑_{|w| = |u|} ∑_{|w'| = |v|} ⟨u|Γ(U†)|w⟩ D(z)_{ww'} ⟨w'|Γ(U)|v⟩ = D(U†z)_{uv}`.
**Formalization Note.** `Γ` preserves the total photon number, so both sums are finite; they are written over the
box `w_i ≤ |u|` filtered by `|w| = |u|`, as in `gammaFock_mul`. `D(z)_{mn}` is `Dependency50.P39.displacementMatrix`. -/
theorem gammaFock_weyl {n : ℕ} (U : Matrix (Fin n) (Fin n) ℂ) (hU : Uᴴ * U = 1) (z : Fin n → ℂ)
(u v : Occupation (Fin n)) :
∑ w ∈ (Fintype.piFinset fun _ : Fin n => Finset.range (∑ i, u i + 1)).filter
(fun w => ∑ i, w i = ∑ i, u i),
∑ w' ∈ (Fintype.piFinset fun _ : Fin n => Finset.range (∑ i, v i + 1)).filter
(fun w' => ∑ i, w' i = ∑ i, v i),
gammaFock Uᴴ u w * (∏ j, displacementMatrix (z j) (w j) (w' j)) * gammaFock U w' v =
∏ j, displacementMatrix ((Uᴴ *ᵥ z) j) (u j) (v j) := by sorry
end BosonicLearn
Source
Chen, Anna Mele, Anna Mele, Preskill, When are bosonic Gaussian states classical to learn?, arXiv:2609.26705, §1.1, Eqs. (2) and (4), pp. 4–5 (Gaussian states, characteristic function, passive Gaussian states); Lemma 2 and Eq. (21), p. 20 (passive Gaussian states as det(I − R)Γ(R))