For Banach spaces and , define the weak -norm of a finite family byA bounded linear operator is an absolutely p-summing operator if for every finite family. The least constant is .
The Pietsch factorization theorem states that such an operator admits a probability measure on with its weak-star topology such thatEquivalently, let be the evaluation embedding and the natural map. There is a bounded linear operator , where , such that and . Conversely, any such factorization makes an absolutely p-summing operator, with . The final operator is required on this closed subspace ; no extension to all of is assumed.
Here is the existence proof. If the conclusion is immediate. Otherwise write . The Banach-Alaoglu theorem makes compact, and is continuous for each . For finitely many vectors , putWe claim the compact convex hull of meets . If it did not, finite-dimensional Hahn-Banach separation theorem would give weights withThe weights are nonnegative because the second convex set is unbounded in every positive coordinate direction. This strict inequality contradicts the defining summing inequality applied to . Thus a finite convex combination of the vectors lies coordinatewise above . Equivalently, a finitely supported probability measure satisfies all these finitely many domination constraints.
By the Riesz-Markov-Kakutani representation theorem, probability measures on compact are the positive norm-one functionals on taking the constant function to one. They form a compact set in the weak-star topology, by the Banach-Alaoglu theorem. Each domination constraint is closed. The finite feasibility just proved gives the finite intersection property, hence one measure satisfies every constraint.
Define . The domination inequality makes this well-defined and bounded by , and completeness of extends it to . Conversely, any factorization satisfiesThis proves both directions of the Pietsch factorization theorem, including the optimal factorization norm.
For the multiplication operator, let and take . Point evaluations are norm-one linear functionals on , soThus is an absolutely p-summing operator and . Taking the single function gives the reverse inequality, soFor , the explicit domination measure on is , pushed into by . For everything is zero; no essential boundedness of is needed.
Articles by others on the same topic
There are currently no matching articles.