Definition

A locally convex space is a real or complex EE for which 00 has a basis of . Equivalently, its topology is generated by a separating family of . The neighborhoods may be chosen both convex and balanced. Under this corpus's convention, topological vector spaces are Hausdorff, so Hausdorffness is included here. Local convexity supports separation and duality while allowing topologies not defined by one norm.

Seminorm description

For a family P\mathcal P of seminorms on EE, typical basic zero-neighborhoods have the form

U(p1,,pn;ε)={xE:pj(x)<ε for 1jn},U(p_1,\ldots,p_n;\varepsilon) = \{x\in E:p_j(x)<\varepsilon\text{ for }1\leq j\leq n\},

where p1,,pnPp_1,\ldots,p_n\in\mathcal P and ε>0\varepsilon>0. Conversely, the Minkowski gauges of suitable convex balanced zero-neighborhoods recover defining seminorms. This correspondence is formalized by the .

Separation

The seminorm topology is Hausdorff exactly when the defining family : for each nonzero xEx\in E, some pPp\in\mathcal P satisfies p(x)>0p(x)>0. Under this separation hypothesis, the Hahn–Banach theorem gives a rich and makes continuous linear functionals effective probes of the space.

Examples and nonexamples

Every is locally convex, using its norm as one seminorm. , spaces of smooth functions, and naturally require families of seminorms. By contrast, many LpL^p-spaces with 0<p<10<p<1, equipped with their usual complete translation-invariant metrics, are topological vector spaces whose topologies are not locally convex.

The seminorm formulation and its use in distribution theory are developed in Trèves, Chapter 7.

References
  1. François Trèves, Topological Vector Spaces, Distributions and Kernels, Academic Press, 1967. Elsevier publisher record. Relevant: Chapter 7.
  2. Helmut H. Schaefer and Manfred P. Wolff, Topological Vector Spaces, 2nd ed., Springer, 1999. Springer DOI record. Relevant: Chapter II, “Locally Convex Topological Vector Spaces.”