Definition
Nuclear space
A locally convex space whose defining Banach-space transition maps are nuclear.
Definition
Let be a Hausdorff locally convex space. For a continuous seminorm , let be the Banach completion of . The space is nuclear if, for every continuous seminorm , there is a continuous seminorm such that the canonical map
is a nuclear operator: it has an absolutely summable rank-one decomposition. This is a property of the locally convex topology, not merely of the underlying vector space or its topological dual as a set.
Why the definition is strong
The transition map condition forces finite-dimensional-like summability between successively stronger seminorms. It implies that the projective and injective locally convex tensor-product constructions with agree after the appropriate Hausdorff completion. This removes a major ambiguity in kernel theorems and makes continuous multilinear maps unusually tractable Trèves, Chapters 50–51.
Examples and nonexamples
Finite-dimensional locally convex spaces are nuclear. The Schwartz space , the test-function space , and the space on a compact smooth manifold are fundamental infinite-dimensional examples with their standard locally convex topologies. By contrast, a Banach space is nuclear as a locally convex space only when it is finite-dimensional: applying the definition to a norm would make an identity-type transition map nuclear, hence compact.
Topological cautions
Nuclearity does not mean that every element is a rapidly decreasing function, nor does it define a class of distributions. Those are properties of particular function spaces and their continuous duals. Changing the topology on a fixed vector space can change whether it is nuclear, because continuity of the seminorms and nuclearity of the completion maps both depend on that topology Schaefer–Wolff, Chapter IV.
References
- François Trèves, Topological Vector Spaces, Distributions and Kernels, Academic Press, 1967. Publisher record. Relevant: Chapters 50–51 on nuclear spaces and kernel theorems.
- Helmut H. Schaefer and Manfred P. Wolff, Topological Vector Spaces, 2nd ed., Springer, 1999. Publisher record. Relevant: Chapter IV on nuclear maps and spaces.