Threshold separation for sums of independent geometric variables: (Lemmas 13–14)
ProvedBosonicLearn.geometric_pgf_thresholdLemmas 13 and 14 of the paper, combined and specialized to the photon-count distributions of the proof of Lemma 12.
Fix . There is a constant with the following property. Let for , and let
be sums of independent geometric random variables on with means (so ). Put
If , then there is a threshold such that
Role. In the proof of Lemma 12, are the photon counts in a subspace of modes for the two passive Gaussian states; the linear dependence on (rather than the quadratic one a second-moment argument would give) is what yields the trace-distance bound .
Formalization Note The two laws are sequences pp pm : ℕ → ℝ with non-negative entries, specified through their probability generating functions on (HasSum (fun k => p k * z ^ k) (∏ i, (1 + λ i * (1 - z))⁻¹)), which determine them; at this says they have total mass one. The hypothesis is written . The constant depends only on (the proof gives ).
import Mathlib
namespace BosonicLearn
/-- **Threshold separation for sums of independent geometric variables (Lemmas 13 and 14 combined).**
Chen, Anna Mele, Anna Mele, Preskill, *When are bosonic Gaussian states classical to learn?*, arXiv:2609.26705, §5.7, Lemma 13 (coarse Lévy concentration bound), Lemma 14 (threshold separation under convolution order), and the coupling
of Eqs. (137)–(142).
Fix `Λ > 0`. There is `c > 0` such that the following holds. Let `0 ≤ λ⁻_i ≤ λ⁺_i ≤ Λ` (`i = 1, …, m`), and let
`p⁺, p⁻` be the laws on `ℕ` of `K₊ = ∑ᵢ M⁺ᵢ` and `K₋ = ∑ᵢ M⁻ᵢ`, sums of independent geometric variables with
means `λ^±_i` — given through their probability generating functions
`∑ₖ p^±_k zᵏ = ∏ᵢ 1/(1 + λ^±_i (1 − z))`, `0 ≤ z ≤ 1`. Put `δ = ∑ᵢ (λ⁺_i − λ⁻_i)` and
`s² = 1 + Var(K₋) = 1 + ∑ᵢ λ⁻_i(1 + λ⁻_i)`. If `0 < δ ≤ s`, then there is a threshold `τ` with
`Pr(K₊ > τ) − Pr(K₋ > τ) ≥ c δ / s`.
**Formalization Note.** The laws are sequences `ℕ → ℝ` with non-negative entries, determined by the generating
function identities (at `z = 1` they give total mass one). `δ ≤ s` is written `δ² ≤ s²`. -/
theorem geometric_pgf_threshold :
∀ Λ : ℝ, 0 < Λ → ∃ c : ℝ, 0 < c ∧
∀ (m : ℕ) (lp lm : Fin m → ℝ), (∀ i, 0 ≤ lm i) → (∀ i, lm i ≤ lp i) → (∀ i, lp i ≤ Λ) →
∀ (pp pm : ℕ → ℝ), (∀ k, 0 ≤ pp k) → (∀ k, 0 ≤ pm k) →
(∀ z : ℝ, 0 ≤ z → z ≤ 1 →
HasSum (fun k => pp k * z ^ k) (∏ i, (1 + lp i * (1 - z))⁻¹)) →
(∀ z : ℝ, 0 ≤ z → z ≤ 1 →
HasSum (fun k => pm k * z ^ k) (∏ i, (1 + lm i * (1 - z))⁻¹)) →
0 < ∑ i, (lp i - lm i) →
(∑ i, (lp i - lm i)) ^ 2 ≤ 1 + ∑ i, lm i * (1 + lm i) →
∃ τ : ℕ, c * (∑ i, (lp i - lm i)) / Real.sqrt (1 + ∑ i, lm i * (1 + lm i)) ≤
(∑' k, if τ < k then pp k else 0) - ∑' k, if τ < k then pm k else 0 := by sorry
end BosonicLearn