Finite-rank approximation on bounded sets for maps from nuclear spaces
ProvedMasulloGRH.nuclear_finite_rank_approximationLet be a nuclear Hausdorff locally convex complex vector space, a continuous complex-linear map, a bounded set, a continuous seminorm on , and . There are finitely many continuous linear functionals and vectors such that
Thus continuous maps from nuclear spaces admit finite-rank approximations uniformly on a prescribed bounded set in each continuous target seminorm. Taking to be the identity gives the corresponding approximation property.
This is the neighborhood-estimate formulation of the finite-rank density assertion in Remark C.2.4.17. Finite families of target seminorms and bounded sets can be absorbed into a maximum seminorm and finite union, respectively. No completeness, metrizability, barrelledness, or quasi-completeness assumption on is required.
Formalization Note: the statement records the approximants explicitly as finite sums of continuous rank-one maps. It does not package them as a separate dense subset of an operator-topology type; the target is only required to carry the displayed vector-space structure, topology, continuous map, and continuous seminorm.
import Definitions.Def_MasulloGRH_LocallyConvex set_option autoImplicit false open MasulloGRH Filter Topology Set open scoped BigOperators
theorem MasulloGRH.nuclear_finite_rank_approximation (E F : Type*) [AddCommGroup E] [Module ℂ E] [TopologicalSpace E]
[AddCommGroup F] [Module ℂ F] [TopologicalSpace F]
(hE : IsNuclear E) (T : E →L[ℂ] F) (B : Set E)
(hB : Bornology.IsVonNBounded ℂ B) (q : Seminorm ℂ F) (hq : Continuous q)
(ε : ℝ) (hε : 0 < ε) :
∃ (N : ℕ) (φ : Fin N → StrongDual ℂ E) (y : Fin N → F),
∀ x ∈ B, q (T x - ∑ i, φ i x • y i) < ε := by sorry