Closed graph theorem (functional analysis)

In mathematics, particularly in functional analysis and topology, the closed graph theorem is a fundamental result stating that a linear operator with a closed graph will, under certain conditions, be continuous. The original result has been generalized many times so there are now many theorems referred to as "closed graph theorems."

DefinitionsEdit

Graphs and closed graphsEdit

Definition and notation: The graph of a function f : XY is the set
Gr f := { (x, f(x)) : xX } = { (x, y) ∈ X × Y : y = f(x) }.
Assumption: If X and Y are topological spaces then X × Y will always be endowed with the product topology.
Definition: If X and Y are topological spaces, DX, and f : DY is a function, then we say that f  has a closed graph (resp. sequentially closed graph) in X × Y if the graph of f, Gr f, is a closed (resp. sequentially closed) subset of X × Y. If D = X or if X is clear from context then we may omit writing "in X × Y"

Linear operatorsEdit

Definition and notation:[1] A partial map, denoted by f : XY, if a map from a subset of X, denoted by dom f, into Y. If we write f : DXY then we mean that f : XY is a partial map and dom f = D.
Definition and notation: If we say that f : DXY is closed (resp. sequentially closed) or has a closed graph (resp. has a sequentially closed graph) then we mean that the graph of f is closed (resp. sequentially closed) in X × Y (rather than in D × Y).
Definition: If we say that f : DXY is a linear or a linear operator then we mean that X and Y are vector spaces, DX is a vector subspace of X, f : DY is a linear map.

Closed linear operatorsEdit

Assumption: We will henceforth assume that X and Y are topological vector spaces (TVSs).
Definition: A linear operator f : DXY is called closed or a closed linear operator if its graph is closed in X × Y.
Closable maps and closures
Definition: We say that a linear operator f : DXY is closable in X × Y if there exists a vector subspace EX containing S and a function (resp. multifunction) F : EY whose graph is equal to the closure of the set Gr f in X × Y. Such an F is called a closure of f in X × Y, is denoted by f, and necessarily extends f.
Definition: If f : DXY is closable linear operator then a core or essential domain of f is a subset CD such that the closure in X × Y of the graph of the restriction f|C : CY of f to C is equal to the closure of the graph of f in X × Y (i.e. the closure of Gr f in X × Y is equal to the closure of Gr f|C in X × Y).
Closed maps vs. closed linear operators

When reading literature in functional analysis, if f : XY is a linear map between topological vector spaces (TVSs) then "f is closed" will almost always mean that its graph is closed. However, "f is closed" may, especially in literature about point-set topology, instead mean the following:

Definition: A map f : XY between topological spaces is called a closed map if the image of a closed subset of X is a closed subset of Y.

These two definitions of "closed map" are not equivalent. If it is unclear, then it is recommended that a reader check how "closed map" is defined by the literature they are reading.

Characterizations of closed graphs (general topology)Edit

Throughout, let X and Y be topological spaces.

Function with a closed graph

If f : XY is a function then the following are equivalent:

  1. f  has a closed graph (in X × Y);
  2. (definition) the graph of f, Gr f, is a closed subset of X × Y;
  3. for every xX and net x = (xi)iI in X such that xx in X, if yY is such that the net f(x) := (f(xi))iIy in Y then y = f(x);[2]
    • Compare this to the definition of continuity in terms of nets, which recall is the following: for every xX and net x = (xi)iI in X such that xx in X, f(x) → f(x) in Y.
    • Thus to show that the function f has a closed graph we may assume that f(x) converges in Y to some yY (and then show that y = f(x)) while to show that f is continuous we may not assume that f(x) converges in Y to some yY and we must instead prove that this is true (and moreover, we must more specifically prove that f(x) converges to f(x) in Y).

and if Y is a Hausdorff compact space then we may add to this list:

  1. f  is continuous;[3]

and if both X and Y are first-countable spaces then we may add to this list:

  1. f  has a sequentially closed graph (in X × Y);
Function with a sequentially closed graph

If f : XY is a function then the following are equivalent:

  1. f  has a sequentially closed graph (in X × Y);
  2. (definition) the graph of f is a sequentially closed subset of X × Y;
  3. for every xX and sequence x = (xi)
    i=1
    in X such that xx in X, if yY is such that the net f(x) := (f(xi))
    i=1
    y
    in Y then y = f(x);[2]

Basic properties of maps with closed graphsEdit

The following properties are easily checked for a linear operator f : D(f) ⊆ XY between Banach spaces:

  • If A is closed then AλIdD(f) is closed where λ is a scalar and IdD(f) is the identity function;
  • If f is closed, then its kernel (or nullspace) is a closed vector subspace of X;
  • If f is closed and injective then its inverse f−1 is also closed;
  • A linear operator f admits a closure if and only if for every xX and every pair of sequences x = (xi)
    i=1
    and y = (yi)
    i=1
    in D(f) both converging to x in X, such that both f(x) = (f(xi))
    i=1
    and f(y) = (f(yi))
    i=1
    converge in Y, one has limi → ∞ fxi = limi → ∞ fyi.

Examples and counterexamplesEdit

Continuous but not closed mapsEdit

  • Let X denote the real numbers with the usual Euclidean topology and let Y denote with the indiscrete topology (where note that Y is not Hausdorff and that every function valued in Y is continuous). Let f : XY be defined by f(0) = 1 and f(x) = 0 for all x ≠ 0. Then f : XY is continuous but its graph is not closed in X × Y.[2]
  • If X is any space then the identity map Id : XX is continuous but its graph, which is the diagonal Gr Id := { (x, x) : xX}, is closed in X × X if and only if X is Hausdorff.[4] In particular, if X is not Hausdorff then Id : XX is continuous but not closed.
  • If f : XY is a continuous map whose graph is not closed then Y is not a Hausdorff space.

Closed but not continuous mapsEdit

  • Let (X, 𝜏) be a Hausdorff TVS and let 𝜐 be a vector topology on X that is strictly finer than 𝜏. Then the identity map Id : (X, 𝜏) → (X, 𝜐) a closed discontinuous linear operator.[5]
  • Consider the derivative operator A = d/dx where X = Y = C([a, b]) is the Banach space of all continuous functions on an interval [a, b]. If one takes its domain D(f) to be C1([a, b]), then f is a closed operator, which is not bounded.[6] On the other hand if D(f) = C([a, b]), then f will no longer be closed, but it will be closable, with the closure being its extension defined on C1([a, b]).
  • Let X and Y both denote the real numbers with the usual Euclidean topology. Let f : XY be defined by f(0) = 0 and f(x) = 1/x for all x ≠ 0. Then f : XY has a closed graph (and a sequentially closed graph) in X × Y = ℝ2 but it is not continuous (since it has a discontinuity at x = 0).[2]
  • Let X denote the real numbers with the usual Euclidean topology, let Y denote with the discrete topology, and let Id : XY be the identity map (i.e. Id(x) := x for every xX). Then Id : XY is a linear map whose graph is closed in X × Y but it is clearly not continuous (since singleton sets are open in Y but not in X).[2]

Closed graph theoremsEdit

The closed graph theorem can be generalized from Banach spaces to more abstract topological vector spaces in the following ways:

Theorem — A linear operator from a barrelled space X to a Fréchet space Y is continuous if and only if its graph is closed.

and there are versions that does not require Y to be locally convex:

Theorem[7][8] — A linear map between two F-spaces is continuous if and only if its graph is closed.

We restate this theorem and extend it with some conditions that can be used to determine if a graph is closed:

Theorem — If T : XY is a linear map between two F-spaces, then the following are equivalent:

  1. T is continuous;
  2. T has a closed graph;
  3. If (xi)
    i=1
    x
    in X and if (T(xi))
    i=1
    converges in Y to some yY, then y = T(x);[9]
  4. If (xi)
    i=1
    → 0
    in X and if (T(xi))
    i=1
    converges in Y to some yY, then y = 0

Theorem[10] — Suppose that T : XY is a linear map whose graph is closed. If X is an inductive limit of Baire TVS and Y is a webbed space then T is continuous.

Closed Graph Theorem[11] — A closed surjective linear map from a complete pseudometrizable TVS onto a locally convex ultrabarrelled space is continuous.

Also, a closed linear map from a locally convex ultrabarrelled space into a complete pseudometrizable TVS is continuous.

Closed Graph Theorem[11] — A closed and bounded linear map from a locally convex infrabarreled space into a complete pseudometrizable locally convex space is continuous.

An even more general version of the closed graph theorem is

Theorem[12] — Suppose that X and Y are two topological vector spaces (they need not be Hausdorff or locally convex) with the following property: If G is any closed subspace of X × Y and u is any continuous map of G onto X, then u is an open mapping. Under this condition, if T : XY is a linear map whose graph is closed then T is continuous.

Between Banach spacesEdit

In functional analysis, the closed graph theorem states the following: If X and Y are Banach spaces, and T : XY is a linear operator, then T is continuous if and only if its graph is closed in X × Y (with the product topology).

The closed graph theorem can be reformulated may be rewritten into a form that is more easily usable:

Closed Graph Theorem for Banach spaces — If T : XY is a linear operator between Banach spaces, then the following are equivalent:

  1. T is continuous.
  2. T is a closed operator (i.e. the graph of T is closed).
  3. If (xi)
    i=1
    x
    in X then (T(xi))
    i=1
    T(x)
    in Y.
  4. If (xi)
    i=1
    → 0
    in X then (T(xi))
    i=1
    → 0
    in Y.
  5. If (xi)
    i=1
    x
    in X and if (T(xi))
    i=1
    converges in Y to some yY, then y = T(x).
  6. If   in X and if (T(xi))
    i=1
    converges in Y to some yY, then y = 0.

Note that the operator is required to be everywhere-defined, that is, the domain D(T) of T is X. This condition is necessary, as there exist closed linear operators that are unbounded (not continuous); a prototypical example is provided by the derivative operator on C([0,1]), whose domain is a strict subset of C([0,1]).

The usual proof of the closed graph theorem employs the open mapping theorem. In fact, the closed graph theorem, the open mapping theorem and the bounded inverse theorem are all equivalent. This equivalence also serves to demonstrate the importance of X and Y being Banach; one can construct linear maps that have unbounded inverses in this setting, for example, by using either continuous functions with compact support or by using sequences with finitely many non-zero terms along with the supremum norm.

Borel graph theoremEdit

The Borel graph theorem, proved by L. Schwartz, shows that the closed graph theorem is valid for linear maps defined on and valued in most spaces encountered in analysis.[13] Recall that a topological space is called a Polish space if it is a separable complete metrizable space and that a Souslin space is the continuous image of a Polish space. The weak dual of a separable Fréchet space and the strong dual of a separable Fréchet-Montel space are Souslin spaces. Also, the space of distributions and all Lp-spaces over open subsets of Euclidean space as well as many other spaces that occur in analysis are Souslin spaces. The Borel graph theorem states:[13]

Borel Graph Theorem — Let u : XY be linear map between two locally convex Hausdorff spaces X and Y. If X is the inductive limit of an arbitrary family of Banach spaces, if Y is a Souslin space, and if the graph of u is a Borel set in X × Y, then u is continuous.

An improvement upon this theorem, proved by A. Martineau, uses K-analytic spaces.

Definition: A topological space X is called a Kσδ if it is the countable intersection of countable unions of compact sets.
Definition: A Hausdorff topological space Y is called K-analytic if it is the continuous image of a Kσδ space (that is, if there is a Kσδ space X and a continuous map of X onto Y).

Every compact set is K-analytic so that there are non-separable K-analytic spaces. Also, every Polish, Souslin, and reflexive Frechet space is K-analytic as is the weak dual of a Frechet space. The generalized Borel graph theorem states:

Generalized Borel Graph Theorem[14] — Let u : XY be a linear map between two locally convex Hausdorff spaces X and Y. If X is the inductive limit of an arbitrary family of Banach spaces, if Y is a K-analytic space, and if the graph of u is closed in X × Y, then u is continuous.

Related resultsEdit

Theorem[15] — If F : XY is closed linear operator from a Hausdorff locally convex TVS X into a Hausdorff finite-dimensional TVS Y then F is continuous.


See alsoEdit

ReferencesEdit

  1. ^ Dolecki & Mynard 2016, pp. 4-5.
  2. ^ a b c d e Narici & Beckenstein 2011, pp. 459-483.
  3. ^ Munkres 2000, p. 171.
  4. ^ Rudin 1991, p. 50.
  5. ^ Narici & Beckenstein 2011, p. 480.
  6. ^ Kreyszig, Erwin (1978). Introductory Functional Analysis With Applications. USA: John Wiley & Sons. Inc. p. 294. ISBN 0-471-50731-8.
  7. ^ Schaefer & Wolff 1999, p. 78.
  8. ^ Trèves (1995), p. 173
  9. ^ Rudin 1991, pp. 50-52.
  10. ^ Narici & Beckenstein 2011, p. 479-483.
  11. ^ a b Narici & Beckenstein 2011, pp. 474-476.
  12. ^ Trèves 2006, p. 169.
  13. ^ a b Trèves 2006, p. 549.
  14. ^ Trèves 2006, pp. 557-558.
  15. ^ Narici & Beckenstein 2011, p. 476.

BibliographyEdit