Roughly speaking, Minkowski functionals are like seminorms, but more general and sis defined associated to some subsets of a topological vector space. Under some conditions, it can even be norms, making the topological vector space a normed space. To extend this idea, we will also discuss locally convex topological vector spaces, which are better than general topological vector spaces, and can be described by a separating family of seminorms.

SpacesStructures that Determine the Topology
Topological Vector SpacesA family of Minkowski functionals
Locally Convex SpacesA family of Seminorms
Normable SpacesA norm

Minkowski Functional

Suppose is a subset of a real or complex vector space. Then the Minkowski functional or gauge of is defined to be a function such that

Conventionally, we set the infimum over an empty set to be .

e.g.

  • Consider a normed vector space and let be the unit ball in . Then the Minkowski functional is just the norm on .

Proposition

If is convex, then is bounded;

Seminorms

Seminorm

A seminorm on a complex or real topological vector space is a map such that the following properties hold:

  • nonnegativity: for all ,
  • absolute homogeneity: for all and ,
  • triangle inequality: for all .

Proposition

Suppose is a seminorm on a real or complex vector space . Then the -ball w.r.t , is convex, balanced, and absorbing.

Proposition

Suppose is a family of seminorms on a real or complex vector space . Then the initial topology generated by (i.e., the topology generated by neighbourhood subbasis at ) is a locally convex topology on .

Proposition

Let be a vector space, and let be a family of seminorms on . Then the locally convex topology generated by is Hausdorff if and only if .

Locally Convex Spaces

Local Convexity

A topological vector space is locally convex if it has a basis of its topology consisting of convex open subsets. Equivalently, has a neighborhood basis consisting of convex open subsets.

Lemma

A complete topological vector space is complete in every compatible metric that induces its topology.

Fréchet Space

A topological vector space is a Fréchet space if it is, locally convex, metrizable and complete.

Remark

The original definition due to Stefan Banach defines Fréchet spaces as metrizable complete topological vector spaces, which does not require local convexity. The books of Waelbroeck (1971) and Wilansky (1978) follow this original definition.

Alternative Characterization of Locally Convex Spaces

A topological vector space is locally convex if and only if its topology is induced by a countable family of seminorms. That is, its topology is the coarsest topology (i.e., initial topology) such that each seminorm is continuous.

The following propositions help to verify if a given space is Fréchet:

Proposition

A locally convex space is metrizable if and only if its topology can be generated by a countable separating (for every non-zero , there must be at least one seminorm ​ such that ) family of seminorms.

e.g.

  • Clearly, any Banach space is a Fréchet space, since it is complete and metrizable by the norm.
  • The space of real smooth functions is a Fréchet space with the seminorms defined by for each nonnegative integers and .
  • If is a compact smooth manifold, then the space of smooth functions is a Fréchet space. This holds for some non-compact manifolds as well, such as the real line .