Complemented subspace

In the branch of mathematics called functional analysis, a complemented subspace[note 1] of a normed space or (more generally) of a topological vector space is a vector subspace for which there exists some other vector subspace of called its (topological) complement in that allows to be treated as if it were the direct sum or the product both algebraically and topologically. Explicitly, this means that is the algebraic direct sum of and and also that the addition map that sends is an topological vector space isomorphism (or equivalently, a homeomorphism), which is a situation is summarized by saying that is the topological direct sum of and [1] or that is their direct sum in the category of topological vector spaces.

If is the direct sum of and then this is expressed by writing although it may need to be clarified whether this is merely the direct sum in the category of vector spaces (that is, an algebraic direct sum, which is true if and only if and ) or the direct sum in the category of topological vector spaces (that is, a topological direct sum). Every topological direct sum is an algebraic direct sum but the converse is not guaranteed in general. This is because although the aforementioned addition map is always continuous, even if both and are closed in and is also the algebraic direct sum of and (in which case this continuous linear addition map is a bijection), this addition map might nevertheless still fail to be a homeomorphism, or said differently, might not be a topological complement of [1] However, even if this particular is not a topological complement of it might still be possible for some other vector subspace to be 's topological complement. If has no topological complement in then it is said to be uncomplemented[note 1] in The property of "being a topological complement of" is symmetric in the sense that is a topological complement of if and only if is a topological complement of

A complemented (vector) subspace of a Hausdorff space is necessarily a closed subset of [1][proof 1] (and so the same is thus also true of its complement[note 2]), which is why most definitions of "complemented subspace" include the requirement that it (and possibly also its complement) be closed in However, even if being closed in is not mentioned as part of the definition, it will nevertheless still be a necessary property that any complemented subspace (and its complement) must possess. In particular, every vector subspace of a Hausdorff space that is not closed is necessarily uncomplemented, which is why only closed subspaces are typically considered. Even considering only closed subspaces, the definition of a "complemented subspace" would still be needed because it is known that there exist spaces that possess closed uncomplemented subspaces. No Hilbert spaces (and so in particular, no finite-dimensional Euclidean space) has any closed uncomplemented subspace because the orthogonal complement of any closed vector subspace is always a topological complement of In fact, a Banach space is a Hilbert space if and only if every closed vector subspace is complemented. In particular, an arbitrary closed vector subspace of a non-Hilbert Banach space is not guaranteed to be complemented.

The concept of a complemented subspace in functional analysis should never be confused with that of a set complement in set theory, which is completely different.

Preliminaries

Algebraic direct sum

If is a vector space and and are vector subspaces of then the addition map

is a linear map or said differently, it is a morphism in the category of vector spaces. The vector space is said to be the algebraic direct sum of and or the direct sum of and in the category of vector spaces if any of the following equivalent conditions are satisfied:

  1. The addition map is a vector space isomorphism.[1][2]
  2. The addition map is bijective.
  3. and ; in this case is called an algebraic complements or an algebraic supplements to in and these spaces are said to be complementary or supplementary.[1]

The condition can be expressed by saying that and are linearly independent vector subspaces.[3] More generally, a collection of subspaces of are said to be linearly independent if for every index where [3]

Motivation

Suppose that the vector space is the algebraic direct sum of in Then algebraically, may be treated as if it were or Assuming now that is also a topological vector space, such as any normed space, then the addition map

will necessarily be a continuous[note 3] bijection but its inverse may fail to be continuous. However, if (and only if) this map is continuous will be the topological direct sum of and which by definition is just another way of say that is the (categorical) direct sum of and in the category of topological vector spaces (rather than merely the direct sum in the category of vector spaces). This is equivalent to the addition map being a linear homeomorphism, which when true, allows for to be treated as if it were or both algebraically and topologically. That this is not always true is the reason why definitions such as "complemented subspace" are needed, where a subspace is called (topologically) complement in if there exists some other subspace that allows to be treated as if it were or both algebraically and topologically.

For an example where is an algebraic, but not topological, direct sum of two vector subspaces and suppose is a discontinuous linear functional on a normed space (where a linear functional is discontinuous if and only if is not closed in ) and let be such that Then is the algebraic direct sum of and but is not the topological direct sum of and because is not closed in This subspace being closed is necessary when is Hausdorff[1] because then is closed in the product topology on so that being a homeomorphism would force to be closed in (and similarly for ). This proof generalized to show that a topologically complemented subspace of a Hausdorff topological vector space (TVS) must be closed in which is why many authors require this property (that is, being closed) as part of the definition of a complemented subspace. This counter example also generalizes to any Hausdorff locally convex TVS that admits a discontinuous linear functional and moreover, the kernel of such a discontinuous linear functional cannot be topologically complemented in (because it is not a closed subspace of ).

Because a linear map between two normed (or Banach) spaces is bounded if and only if it is continuous, if is a normed space (resp. a Banach space) then is the direct sum of and in the category of normed spaces (resp. the category of Banach spaces) if and only if is the topological direct sum of and

Notation: Inverse of addition and canonical projections

If is the algebraic direct sum of and then by definition, the addition map

is a linear bijection so its inverse is well-defined and it can be written in terms of coordinates as

where the first coordinate is called the canonical projection of onto while the second coordinate is called the canonical projection of onto [2] To clarify, for any the defining property of and is that they are the unique vectors in and respectively, that satisfy

These maps satisfy

where denotes the identity map on

Definitions

As above, let denote the addition map and when is injective, then let be its inverse.

Topological direct sum

If is a topological vector space (TVS) and if and are vector subspaces of then is the topological direct sum of and if any of the following equivalent conditions are satisfied:

  1. The addition map is a TVS-isomorphism (that is, a surjective linear homeomorphism).[1]
  2. is the (categorical) direct sum of and in the category of topological vector spaces (whose morphisms are continuous linear operators).
  3. The addition map is a bijective open map.
  4. is the algebraic direct sum of and (that is, the addition map a bijection) and also any of the following equivalent conditions hold:
    1. The inverse of the addition map is continuous;.
    2. Both canonical projections and are continuous.
    3. At least one of the two canonical projections and is continuous.
    4. The map defined by is a TVS-isomorphism.[4]
      • This map is the restriction to of the canonical quotient map defined by The space has quotient topology induced by
    5. The canonical linear map is a bicontinuous vector space isomorphism (meaning and its inverse are continuous);
      • This implies, in particular, that any topological complement of in is TVS-isomorphic to the quotient space
  5. When considered as additive topological groups (so scalar multiplication is ignored), is the topological direct sum of the topological subgroups and

If any of the above conditions hold then and are said to be topological supplements[1] or topological complements in they are said to be (topologically) complementary or supplementary to each other, and each subspace is said to be a (topologically) complemented subspace of

Characterizations of complemented subspaces

Let be a topological vector space and let be a vector subspace of Then is called a complemented subspace of and is said to admit a topological supplement[1] if any of the following equivalent conditions hold:

  1. There exists a vector subspace of such that is the topological direct sum of and (or said differently, is topologically complemented in );
  2. There exists a continuous linear map with image such that ;
    • Such a map is called a continuous linear projection from onto
  3. There exists a continuous linear projection with image such that is the algebraic sum of and the null space of
  4. For every TVS the restriction map is surjective, where is defined by sending a continuous linear operator to its restriction (that is, ).[4]

where if in addition is a Banach space then this list can be extended to include:

  1. is a closed in and there exists another closed subspace of such that is isomorphic to the direct sum via the addition map.

In this case is also a complemented subspace, and and are called complements of each other in

Topological complements of kernels

It can be shown that if is a continuous linear surjection between two TVSs, then the following conditions are equivalent:

  1. The kernel of has a topological complement.
  2. There exists a continuous linear map such that where is the identity map.[4]

Examples and sufficient conditions

If and are TVS then and are topological complements in (since the definition of the product topology implies that the addition map is open). If neither nor are Hausdorff then neither nor is a closed subspace of but they are nevertheless topological complements.[note 4]

If is a TVS then every vector subspace of that is an algebraic complement of is a topological complement of (this is because has the indiscrete topology).[5] Thus every TVS is the product of a Hausdorff TVS and a TVS with the indiscrete topology.[5]

Suppose is a Hausdorff locally convex TVS over the field and is a vector subspace of that is TVS-isomorphic to for some set Then is a closed and complemented vector subspace of (see footnote for proof).[note 5]

Every finite-dimensional vector subspace of a given TVS is topologically complemented if and only if for every non-zero there exists a continuous linear functional on whose value at this point is not zero.[1] Every finite-dimensional vector subspace of a Hausdorff TVS is closed and if the Hausdorff TVS is locally convex then it will be both closed and complemented.[6] A closed finite-codimensional vector subspace of a TVS is complement.[6] A finite-codimensional subspace is not necessarily closed, not even in a Banach space, but if it is closed then it is also complemented. In a Banach or Fréchet space, for example, this could be the case when the finite-codimensional space is the image of a continuous linear map.[7]

If is a proper maximal closed vector subspace of a TVS and if is any algebraic complement of in then is a topological complement of in [8] In particular, if is a non-trivial continuous linear functional on then the kernel of is topologically complemented in

Theorem[9]  Let be a Fréchet space over the field Then the following are equivalent:

  1. does not admit a continuous norm (that is, any continuous seminorm on can not be a norm).
  2. contains a vector subspace that is TVS-isomorphic to
  3. contains a complemented vector subspace that is TVS-isomorphic to

Non-complemented subspaces

Let be a non-normable Fréchet space on which there exists a continuous norm. Then contains a closed vector subspace that has no topological complement.[10]

If is a complete TVS and is a closed vector subspace of such that is not complete, then does not have a topological complement in [10]

Properties

Every closed subspace of a Banach space (or TVS) is complemented if and only if is isomorphic to a Hilbert space. Every topologically complemented vector subspace of a Hausdorff TVS is necessarily closed.[8]

The decomposition method

Theorem[11]  Let and be TVSs such that and Suppose that contains a complemented copy of and contains a complemented copy Then is TVS-isomorphic to

This theorem led to the Schroeder-Bernstein problem: If and are Banach spaces and each is TVS-isomorphic to a complemented subspace of the other, then is TVS-isomorphic to ? In 1996, Gowers provided a negative answer to this question.[11]

Classifying complemented subspaces of a Banach space

Consider a Banach space What are the complemented subspaces of up to isomorphism? Answering this question is one of the complemented subspace problems that remains open for a variety of important Banach spaces, most notably the space (see below).

For some Banach spaces the question is closed. Most famously, if then the only complemented subspaces of are isomorphic to and the same goes for Such spaces are called prime (when their only complemented subspaces are isomorphic to themselves). These are not the only prime spaces, however.

The spaces are not prime whenever in fact, they admit uncountably many non-isomorphic complemented subspaces. The spaces and are isomorphic to and respectively, so they are indeed prime.

The space is not prime due to the fact that it contains a complemented copy of however no other complemented subspaces of are currently known. It is undoubtedly one of the most interesting open problems in functional analysis whether admits other complemented subspaces.

Indecomposable Banach spaces

An infinite-dimensional Banach space is called indecomposable whenever its only complemented subspaces are either finite-dimensional or finite-codimensional. Due to the fact that a finite-codimensional subspace of a Banach space is always isomorphic to that makes indecomposable Banach spaces prime.

The most well-known example of indecomposable spaces are in fact hereditarily indecomposable, which means every infinite-dimensional subspace is also indecomposable. Such spaces are fairly nasty, and have been constructed specifically to defy the sort of behavior we typically desire in a Banach space.

See also

Notes

  1. The reason why the terms "complemented subspace" and "uncomplemented subspace" are not prefixed with the word "topologically" is due to the fact that every vector subspace of any vector space always has some algebraic complement in which is a vector subspace of such that is the algebraic direct sum of and Consequently, the analogous definitions of an "algebraically complemented/uncomplemented vector subspace", were they to be defined, would be useless in functional analysis, which is why these terms are not used. So unlike the term "direct sum", which could refer to either an algebraic or topological direct sum, and the term "complement", which can refer to either an algebraic or topological complement, there is no ambiguity with the terms "complemented" and "uncomplemented".
  2. This is because if is a complement of then is also a complement of in particular, this space is itself complemented (by ) in and so must be closed in
  3. By definition of a topological vector space the addition map must be continuous. The restriction of this continuous map to which is just the map is thus also continuous.
  4. Since and are not Hausdorff, neither is Since the set is not closed in there is no requirement that topological complements in be closed.
  5. Since is a complete topological vector space, so is and since any complete subset of a Hausdorff TVS is closed, is a closed subset of Let be a TVS isomorphism, where each is a continuous linear functional. By the Hahn–Banach theorem, we may extend each to a continuous linear functional on Let so is a continuous linear surjection such that its restriction to is It follows that if we define then the restriction of this continuous linear map is the identity map on so that is a continuous projection onto (that is, ). Thus is complemented in and in the category of TVSs.

Proofs

  1. Assume that is Hausdorff, which implies that is closed in both and and consequently both and are closed subsets of If the addition map defined by is a homeomorphism then both and are closed in

References

  1. Grothendieck 1973, pp. 34–36.
  2. Schaefer & Wolff 1999, pp. 19–24.
  3. Bachman & Narici 2000, pp. 3–7.
  4. Trèves 2006, p. 36.
  5. Wilansky 2013, p. 63.
  6. Rudin 1991, p. 106.
  7. Serre, Jean-Pierre (1955). "Un théoreme de dualité". Commentarii Mathematici Helvetici. 29.1.
  8. Narici & Beckenstein 2011, p. 97.
  9. Jarchow 1981, pp. 129–130.
  10. Schaefer & Wolff 1999, pp. 190–202.
  11. Narici & Beckenstein 2011, pp. 100–101.

Bibliography

This article is issued from Wikipedia. The text is licensed under Creative Commons - Attribution - Sharealike. Additional terms may apply for the media files.