Almost Daugavet centers (Q414799)

A Daugavet centre is a nonzero bounded linear operator between Banach spaces, \(G: X\to Y\), such that \[ \|G+T\|= \|G\|+\|T\| \tag{1} \] for all rank-\(1\) operators \(T=x^*\otimes y\); see [\textit{T.~Bosenko} and \textit{V.~Kadets}, Zh.\ Mat.\ Fiz.\ Anal.\ Geom.\ 6, No.~1, 3--20 (2010; Zbl 1208.46010)]. The special case \(G=\text{Id}_X\) leads, by definition, to the Daugavet property of \(X\) which is studied in many papers. On the other hand, \(X\) is said to have the almost Daugavet property if \[ \|\text{Id}+T\| = 1+\|T\| \] for ``many'' rank-\(1\) operators, meaning for all those \(T=x^*\otimes y\) with \(y\in X\) and \(x^*\) in some norming subspace of \(X^*\); see [\textit{V.~Kadets, V.~Shepelska} and \textit{D.~Werner}, Houston J. Math.\ 37, No.~3, 867--878 (2011; Zbl 1235.46014)]. The paper under review ties these two strands by defining an almost Daugavet centre to be a nonzero \(G:X\to Y\) such that (1) holds for all \(T=x^*\otimes y\) with \(y\in X\) and \(x^*\in G^*(Z)\) for some norming subspace \(Z\) of \(Y^*\). The author gives characterisations of almost Daugavet centres parallel to the characterisations of the almost Daugavet property in [Zbl 1235.46014]; for this he introduces the notion of thickness of an operator. He also proves that for an operator \(G\) to be an almost Daugavet centre up to renorming it is necessary and sufficient that \(G\) fixes a copy of~\(\ell_1\).