Extension of linear operators and Lipschitz maps into \({\mathcal C}(K)\)-spaces (Q2461600): Difference between revisions

In this paper, which is enormous in size, wealth of ideas and technical mastering, the author studies the following problem: Given a bounded linear operator \(T_0: E \to C(K)\), \(E\subset X\) Banach space, does there exist a continuous extension \(T: X\to C(K)\), preferably with good bounds on the norm \(\| T\| \leq\lambda\| T_0\| \)? The same question can be asked about extensions of Lipschitz maps, where the Lipschitz constant plays the role of the norm. The author's key idea is to use the theory of types introduced by \textit{J.-L. Krivine} and \textit{B. Maurey} [Isr. J. Math. 39, 273--295 (1981; Zbl 0504.46013)] and, in particular, the properties \((M)\) and \((M^*)\), introduced by the author in [\textit{N. Kalton}, Ill. J. Math. 37, No. 1, 147--169 (1993; Zbl 0824.46029)] along with their variants \((m_p)\) and \((m_p^*)\) from [\textit{N. Kalton} and \textit{D. Werner}, J. Reine Angew. Math. 461, 137--178 (1995; Zbl 0823.46018)] and the new properties \((L)\) and~\((L^*)\). Here is a sample of the main results. (1) For operators into \(c_0\), for separable \(X\) and \(1< \lambda\leq2\), the linear and the Lipschitz extension problems are equivalent; and it is enough to check pairs \(E\subset F\subset X\) with \(\dim F/E=1\). There is a \(\lambda\)-extension for every subspace \(E\subset X\) if and only if \[ \limsup_{n\to\infty} \| x_n^*\| \leq \limsup_{n\to \infty} \| x^*+x_n^*\| \] whenever \(x_n^*\to 0\) weak* and \(x^*\in X^*\). (2) There are similar, technically more complicated results for operators into~\(c\). (3) The problem of almost isometric extensions (i.e., every \(\lambda>1\) is admissible) can be completely characterised in terms of types. For example, spaces with property~\((M^*)\), like \(\ell_p\) for \(1<p<\infty\) or certain renormings of Orlicz sequence spaces, fall into this category. An example of a particular renorming of \(\ell_2\) shows that almost isometric extensions need not imply isometric extensions (i.e., \(\lambda=1\)). (4) If \(E\subset \ell^1\) is weak* closed, then there are almost isometric extensions into \(C(K)\); this improves an estimate by \textit{W. B. Johnson} and \textit{M. Zippin} [Stud. Math. 117, No. 1, 43--55 (1995; Zbl 0851.46017)] who had obtained \({(3+\varepsilon)}\)-extensions. (5) The first examples of spaces having the following universal property are given: If \(Y\) is embeddable into a separable Banach space \(X\) and \(T_0: Y\to C(K)\) is given, then for any \(\varepsilon>0\) there is an extension \(T: X\to C(K)\) with \(\| T\| \leq (1+\varepsilon)\| T_0\| \). In fact, \(\ell_1\) and indeed the dual of every subspace of \(c_0\) has this property. Again, a characterisation in terms of types is the key point here. (6) The corresponding universal property for Lipschitz extensions is related to a version of the Schur property; under mild additional assumptions it is equivalent to the univeral linear property in (5) for operators \(T:Y\to c\).