Complex interpolation of operators and optimal domains (Q484783)

The complex interpolation of operators from complex interpolation spaces is analyzed from the point of view of the computation of their optimal domains, that is, the maximal Banach lattice of functions to which the operator can be extended. Consider a pair \(X_0\), \(X_1\) of order continuous Banach function spaces over a finite measure space and consider an interpolation pair \((E_0,E_1)\) and an admissible operator \(T:X_0 + X_1 \to E_0 + E_1\) between the pairs \((X_0,X_1)\) and \((E_0,E_1)\). Take the interpolated operator \(T_\theta: [X_0,X_1]_{[\theta]} \to E_0,E_1]_{[\theta]}\) by the first complex interpolation method. On the other hand, consider the optimal domains of the operators \(T|_{X_0}\) and \(T|_{X_1}\), which are given by the spaces \(L^1(m_0)\) and \(L^1(m_1)\) of integrable functions with respect to the corresponding vector measures \(m_0\) and \(m_1\) defined by them, respectively. Take now the optimal domain \(L^1(m_\theta)\) associated to the interpolated operator \(T_\theta\). The paper deals with the problem of finding the relation between \(L^1(m_\theta)\) and the complex interpolation space \([L^1(m_0),L^1(m_1)]_{[\theta]}\). In Theorem 3.1, the authors prove that the space \([L^1(m_0),L^1(m_1)]_{[\theta]}\) is continuously embedded into \(L^1(m_\theta)\). However, they also prove that the reverse inclusion does not hold in general. After this general result, the authors find applications of this framework to analyze two classes of operators that are defined by good factorization properties. The first one is given by the class of the so-called \(p\)-th power factorable operators. In Theorem 4.4, it is shown that the interpolation of two \(p\)-th power factorable operators is again \(p\)-th power factorable. In Theorem 4.5, a general interpolation result is stated: For \(0 < \theta < 1 \leq p_0 \leq p_1 < \infty\), if \(T_0\) is \(p_0\)-th power factorable and \(T_1\) is \(p_1\)-th power factorable, then \(T_{\theta,\alpha}: [X_0,X_1]_{[\theta]} \to [E_0,E_1]_{[\alpha]}\) is well-defined and is a \(p\)-th power factorable restriction of \(T\), where \(p\) is given by \(1/p= 1-\theta/p_0 + \theta/p_1\) and \(\alpha= \theta p /p_1\). As a corollary, the following general result for the interpolation of the class of the \(p\)-th power factorable operators is obtained. Let \( 1 \leq p_0 < p_1 < \infty\) and \(E_0 \subseteq E_1\). If \(T_0\) is \(p_0\)-th power factorable and \(T_1\) is \(p_1\)-th power factorable, then \(T_\theta\) is \(p\)-th power factorable, where \(1/p= 1-\theta/p_0 + \theta/p_1\) (Corollary 4.6). The same results are obtained for the class of the bidual \((p,q)\)-power-concave operators in Theorems 4.9, 4.10 and Corollary 4.11. As an application, it is shown that the following result can be obtained in this setting: If \((X_0,X_1)\) is a pair of \(q\)-convex order continuous Banach function spaces and \(T\) is an admissible operator, and \(T_0\) and \(T_1\) are \(q\)-concave, then so is \(T_\theta\).