Extension operators and twisted sums of \(c_{0}\) and \(C(K)\) spaces (Q1702572): Difference between revisions

The paper [\textit{F. Cabello Sánchez} et al., Trans. Am. Math. Soc. 355, No. 11, 4523--4541 (2003; Zbl 1066.46006)] ended with a comment on a problem that has been in our heads for a time: If \(X\) is a separable Banach space, then \(\mathrm{Ext}(X,c_0)=\{0\}\) by Sobczyk's theorem: we do not know, however, if there is a non-metrizable compact Hausdorff space \(K\) such that \(\mathrm{Ext}(C(K),c_0)=\{0\}\). The problem was followed by a list of cases in which a positive answer could be obtained, which included the statement ``\dots if \(X\) is any non-separable WCG-space, then \(\mathrm{Ext}(X,c_0)\neq\{0\}\) and this settles the case when \(K\) is an Eberlein compact; similar arguments can be used for Corson compact spaces.'' This last assertion about Corson compacta was not right, as \textit{C. Correa} and \textit{D. V. Tausk} [J. Funct. Anal. 270, No. 2, 842--853 (2016; Zbl 1347.46014)] noticed and amended. This paper, in turn, contains further advances on the problem attempting to arrive even at Valdivia compacta. In the meantime, there came all the work about separably injective Banach spaces gathered in [\textit{A. Avilés} et al., Separably injective Banach spaces. Cham: Springer (2016; Zbl 1379.46002)], in which the simplest examples of such spaces, the twisted sums \(Z(c_0, c_0(\Gamma))\) of \(c_0(\Gamma)\) and \(c_0\), were considered. The known examples of spaces \(Z(c_0, c_0(\Gamma))\) are \(C(K)\) spaces, although nobody knows if they have to be. So, a natural question, to the best of my knowledge asked by A. Avilés, was: do those \(C(K)\)-spaces admit nonrivial twisted sums with \(c_0\)? To put a resounding yes to the question, the paper [\textit{J. M. F. Castillo}, Topology Appl. 198, 107--116 (2016; Zbl 1366.46060)] was written. That paper proved more than announced: that any twisted sum space (\(C(K)\) or not) \(Z = Z(c_0(\Gamma'), c_0(\Gamma))\) can be twisted against \(c_0\); i.e., \(\mathrm{Ext}(Z, c_0)\neq 0\) and that the same occurs with all iterations \dots under the continuum hypothesis \mathsf{CH}. The proof goes quite smoothly once the first step has been done and no further trace of \mathsf{CH} appears. And it seems that the presence of \mathsf{CH} is just a technical tool set there simply because the first computation only seems to work for \(\mathfrak c\) and not for \(\aleph_1\). In this scenario, the paper under review appears and the authors show how things actually are: the first step cannot be done for \(\aleph_1\) under different set theoretic assumptions. More precisely, they show that the \(C(K)\) space \(Z(c_0, c_0(\aleph_1))\) cannot be twisted against \(c_0\) under Martin's axiom. And in doing that, they solve the Cabello, Castillo, Kalton, Yost original question. Of course, the paper does much more than that and contains many other significant ideas and advances on the problem of whether/when \(\mathrm{Ext}(C(K), c_0)=0\), but the crucial point is that no study of such questions can now be undertaken outside set theoretic assumptions.