Convexity

A term used in various branches of mathematics and indicating properties which generalize some properties of convex sets (cf. Convex set) in Euclidean spaces . The applicability of a number of methods of study is connected with the term "convexity" .

The following two basic definitions are almost equivalent in . A set is convex: 1) if it is the intersection of open half-spaces; or 2) if it contains with two arbitrary points the segment connecting these points. Both definitions of convexity are applied to the case of vector spaces .

Definition 2) is generalized to include sets in spaces with geodesics (spaces with a connection; locally compact metric spaces, in particular Riemannian and Finsler spaces). The geodesics play the role of segments, but if the two points can be connected by non-unique geodesics or by shortest curves, the concept of convexity branches into several possible meanings. In Riemannian geometry, in particular, the following modifications of convexity are employed [1], [2]:

a set is convex if any two points of can be connected by a unique shortest curve and if this shortest curve is contained in ;

a set is locally convex if each point of has a neighbourhood in which is convex in the sense of ;

a set is weakly convex if any two points can be connected by at least one shortest curve lying in ; and

a set is absolutely convex if all geodesics connecting any two points are located in .

The boundary (or part of the boundary) of an -dimensional convex body in is called a convex hypersurface; if , it is called a convex surface, if , it is said to be a convex curve.

For a function of a real variable convexity signifies the convexity of its supergraph (cf. Convex function (of a real variable)). The convexity of a functional on is defined in a similar manner (cf. Convex functional).

For convex sets in one can speak of the convexity of a family of sets: is convex if provided that , . Convex (and concave) functionals can be defined on convex families . Convexity of a functional is defined by the requirement

The term "convexity" as applied to univalent functions of a complex variable has a special meaning — the property of mapping the unit disc onto a convex domain (cf. Convex function (of a complex variable)).

The -convexity of a compactum , meaning that each point with distance to less than has a unique nearest point in , is one generalization of convexity in which has been studied [4], [5].

In the theory of linear differential operators the term "convexity" is connected with certain properties of homology groups [6]. This is connected with the possibility of touching the boundary from inside a domain by a hypersurface for which a certain number of principal curvatures is positive. In the theory of functions of several complex variables an important role is played by holomorphic convexity, which is connected with the impossibility of touching the boundary of a domain from inside with an analytic surface [7]. The last-named concept is a special case of -convexity [7]. To the concept of -convexity one can transfer many given convexity properties.

The concept of -convexity is employed in convex analysis; it is a generalization of the representability of a convex function as the supremum of a family of linear functions [8].

In the theory of metric spaces, (Menger) convexity of a metric is defined as follows: For any points there exists a third point for which [9]. -convexity of a set means that moreover belongs to if . Definitions of convexity in ordered spaces (cf. Convex subgroup) are very similar.

To almost each definition of convexity there corresponds a concept of local convexity. However, for the class of locally convex topological vector spaces, the term "local convexity" has a special meaning, namely, that for each point there exists a basis system of convex neighbourhoods.

References[edit]

Comments[edit]

Definitions 1) and 2) are equivalent for most, but not all, sets. E.g., a convex set in the sense of 2), maximal with the property of not containing a given point, does not satisfy 1). An explicit example is the space , the unit square minus one half of an edge. The space satisfies 2), but does not satisfy 1).

A more usual definition of convexity of a metric : For any , , and any there is a so that

General results on convexity can be found in [a1], [a6]. Convexity also plays an important role in optimization theory, variational problems and the geometry of numbers (cf. [a2], [a3], [a5], [a7] and [a8]).

For holomorphic convexity cf. Stein manifold; the phrase "K-convex" is also used to denote the property of a set that each connected component is convex ([a4]); cf. Geometry of immersed manifolds for the notion of a -convex surface.

References[edit]