Canonical completion-incidence kernel and spectral gap
ProvedWeldedPath.incidence_gapcompressed-oraclesincidence-operatorsrepresentation-theoryspectral-gap
Let be nonnegative integers with , and let be a vector supported on injective databases of size over an -element set. For an input , define the normalized deletion incidence by
where is the uniform unit vector over the extensions of at (zero when there is no extension). Let , let be the canonical completion map projected to level , and put . Then
and
This identifies the completion space by its domain and image zero sums and bounds leakage from it by incidence energy. The degree-zero case is included.
Preamble
import Mathlib import Definitions.Def_WeldedPath_exactCpO set_option autoImplicit false open Classical open scoped BigOperators
Formal statement
namespace WeldedPath
theorem incidence_gap (s j : ℕ) (hsmall : 3*j ≤ s+1)
(ψ : Hil (InjDb s)) (hsupport : ∀ I : InjDb s, I.size ≠ j → ψ I=0) :
let D : Fin s → Hil (InjDb s) →L[ℂ] Hil (InjDb s) := fun x =>
∑ I ∈ Finset.univ.filter (fun I : InjDb s => I x=none),
ketBra (ket I) (plusState I x)
((∀ x : Fin s, D x ψ=0 ∧ D x (flipDb s ψ)=0) ↔
ψ∈(ContinuousLinearMap.adjoint (Gam s j)).range) ∧
‖ψ-((ContinuousLinearMap.adjoint (Gam s j)).range).starProjection ψ‖^2 ≤
2*∑ x : Fin s, (‖D x ψ‖^2+‖D x (flipDb s ψ)‖^2) := by
sorry
end WeldedPath
Source
Carolan and Majenz, Compressed Permutation Oracles Revisited, https://arxiv.org/html/2609.28469v1, Lemma 4.11 and the incidence Laplacian argument (equation (46)), for the canonical WeldedPath completion operators in https://arxiv.org/html/2611.05912v1, equations (10)-(12). This supporting theorem proves the actual incidence gap and kernel identification; the quantitative one-query comparison is a separate estimate.