Born probabilities are approximated by finite blocks of , uniformly over all effects
ProvedBosonicLearn.born_truncationThe Born probabilities of an infinite-dimensional density operator are approximated by those of its finite blocks, with an error that is uniform over all effects.
Let be an index set and a density operator on (trace class, positive semidefinite, trace one). An effect is a coefficient array with for every finitely supported vector . Write
Then:
- For every there is a finite such that for every finite and every effect ,
- For every finite and every effect , is real and non-negative.
- For every effect , is real and non-negative.
Role. Item 1 is the quantitative form of for the truncation of to the block , with . Because does not depend on , the outcome law of a POVM is a uniform limit of finite-dimensional ones. This is used to build Born laws of POVMs on infinite-dimensional systems (BosonicLearn.fockPOVM_compose), together with BosonicLearn.measure_of_uniform_limit.
Formalization Note The effect condition is the hypothesis effect_form of FockPOVM, on finite sets and coefficient functions : the form has imaginary part and real part in . is the iterated unconditional series ∑' i, ∑' j, E i j * X j i, the form in which FockPOVM.law_born states the Born rule. Density and trace class are IsDensity (nuclear-expansion sense).
Related platform material The space used here (Hilbert A) is the same Lean type as OAI.TensorLp.H A, and is the same type as OAI.EntropyPhotonNumber.Fock n. The published definitions OAI_Analysis_TensorLp (tensor products, slices and operator lifts on ) and OAI_MathematicalPhysics_Fock_Operators (OAI.CrossEnsemble.operator, the bounded operator of a family, with hasSum_operator and entry_hasSum) are sorry-free and can be imported. Proved lemmas of this mission that can be imported: BosonicLearn.traceDist_le_one (nuclear expansions, truncation to finite index sets) and BosonicLearn.effect_matrix_operator (effect matrices as bounded positive operators). The proof imports BosonicLearn.effect_matrix_operator and BosonicLearn.effect_trace_pairing.
import Mathlib import Definitions.Def_BosonicLearn_FockPOVM open QAlgorithms.Papers.Independent.Dependency50.P33 (IsDensity) open scoped ENNReal Matrix
namespace BosonicLearn
/-- **Born probabilities are approximated by finite truncations, uniformly over all effects.**
Chen, Anna Mele, Anna Mele, Preskill, *When are bosonic Gaussian states classical to learn?*, arXiv:2609.26705, §1.1, p. 5 (general POVM measurements; outcome probabilities `Tr(E ρ)`), and Eq. (105), p. 40
(`|Tr(E(ρ − σ))| ≤ ‖ρ − σ‖₁` for `0 ≤ E ≤ I`).
Let `X` be a density operator on `ℓ²(I)`. Call `E` an *effect* if `0 ≤ ⟨z, E z⟩ ≤ ⟨z, z⟩` (real) for every finitely
supported `z`, and write `Tr(E X) = ∑ᵢ ∑ⱼ E_ij X_ji` and `Tr_F(E X) = ∑_{i,j ∈ F} E_ij X_ji` for finite `F ⊆ I`.
1. For every `η > 0` there is a finite `F₀ ⊆ I` such that `|Tr(E X) − Tr_F(E X)| ≤ η` for **every** finite
`F ⊇ F₀` and **every** effect `E`.
2. For every finite `F` and every effect `E`, `Tr_F(E X)` is real and non-negative.
3. For every effect `E`, `Tr(E X)` is real and non-negative.
**Formalization Note.** Effects are coefficient arrays with the bounds of `FockPOVM.effect_form`; density and
trace class are `IsDensity`. The set `F₀` in (1) depends on `X` and `η` only. -/
theorem born_truncation {I : Type} [DecidableEq I] (X : Matrix I I ℂ) (hX : IsDensity X) :
(∀ η : ℝ, 0 < η → ∃ F₀ : Finset I, ∀ F : Finset I, F₀ ⊆ F → ∀ E : Matrix I I ℂ,
(∀ (s : Finset I) (z : I → ℂ),
(∑ i ∈ s, ∑ j ∈ s, star (z i) * E i j * z j).im = 0 ∧
0 ≤ (∑ i ∈ s, ∑ j ∈ s, star (z i) * E i j * z j).re ∧
(∑ i ∈ s, ∑ j ∈ s, star (z i) * E i j * z j).re ≤ ∑ i ∈ s, ‖z i‖ ^ 2) →
‖(∑' i, ∑' j, E i j * X j i) - ∑ i ∈ F, ∑ j ∈ F, E i j * X j i‖ ≤ η) ∧
(∀ (F : Finset I) (E : Matrix I I ℂ),
(∀ (s : Finset I) (z : I → ℂ),
(∑ i ∈ s, ∑ j ∈ s, star (z i) * E i j * z j).im = 0 ∧
0 ≤ (∑ i ∈ s, ∑ j ∈ s, star (z i) * E i j * z j).re ∧
(∑ i ∈ s, ∑ j ∈ s, star (z i) * E i j * z j).re ≤ ∑ i ∈ s, ‖z i‖ ^ 2) →
(∑ i ∈ F, ∑ j ∈ F, E i j * X j i).im = 0 ∧ 0 ≤ (∑ i ∈ F, ∑ j ∈ F, E i j * X j i).re) ∧
(∀ E : Matrix I I ℂ,
(∀ (s : Finset I) (z : I → ℂ),
(∑ i ∈ s, ∑ j ∈ s, star (z i) * E i j * z j).im = 0 ∧
0 ≤ (∑ i ∈ s, ∑ j ∈ s, star (z i) * E i j * z j).re ∧
(∑ i ∈ s, ∑ j ∈ s, star (z i) * E i j * z j).re ≤ ∑ i ∈ s, ‖z i‖ ^ 2) →
(∑' i, ∑' j, E i j * X j i).im = 0 ∧ 0 ≤ (∑' i, ∑' j, E i j * X j i).re) := by sorry
end BosonicLearn