Compact Action Spaces
Definition1
Let and be normed spaces, and let be an operator between these spaces. If for every bounded subset , the image of the operator is precompact, then is called a compact operator.
Explanation
That is precompact means its closure is compact. In other words, a compact operator is an operator that maps bounded sets to precompact sets.
The necessity to study compact operators stems from solving integral equations such as: where is a constant, and the kernel are given functions. The unknown, i.e., the function to be found, is . David Hilbert discovered that the solvability of the above integral equation depends not on the integral form of but solely on the compactness of .
A compact operator is also called a completely continuous operator. This nomenclature comes from the following theorem. Generally, (a) does not imply (b), with (b) being a counterexample.
Theorem
Continuity Theorem: Let and be normed spaces.
(a) Every compact linear operator is bounded. That is, it is continuous.
(b) If is infinite-dimensional, then the identity operator is not compact (though continuous).
Proof
(a)
The unit ball is bounded. Assuming is compact, by the definition of a compact operator, is compact. Being compact implies boundedness, and according to the lemma below, the boundedness of means there exists a such that for all , holds.
The following two propositions are equivalent.
- The subset of the normed space is bounded.
- There exists a positive such that for all , is satisfied.
Therefore,
■
(b)
Assume . The closed ball is bounded. cannot be compact according to Riesz’s lemma. Therefore, is not a compact operator.
For a normed space ,
is finite-dimensional. is compact.
■
Properties
Compact Condition
Let and be normed spaces, and let be a linear operator. Then the following two propositions are equivalent.
- is a compact operator.
- maps “every bounded sequence in ” to “a sequence in that has a converging subsequence”.
Vector Spaces
The set of compact linear operators forms a vector space.
Erwin Kreyszig, Introductory Functional Analysis with Applications (1989), p405-406 ↩︎