Some geometric aspects of operators acting from \(L_{1}\) (Q850600)

There are two main results in this interesting paper. First it is shown that an operator \(T: L_1\to X\) from \(L_1:= L_1[0,1]\) to a Banach space \(X\) is nearly isometric on the positive cone of some \(L_1(A)\)-subspace; more precisely, for every \(\varepsilon>0\) there exists some \(A\subset [0,1]\) such that for all \(f\geq0\) supported on \(A\) one has \(\| Tf\| \geq (\| T\| -\varepsilon)\| f\| \). The second part of the paper deals with a decomposition theorem for the space \(\mathcal L\) of all operators \(T:L_1\to L_1\). A somewhat technical condition \((*)\) of an operator is introduced and it is proved that \(\mathcal L\) decomposes into a direct sum \({\mathcal L} = {\mathcal R} \oplus {\mathcal Z}\) of the subspace \(\mathcal R\) of all representable operators (i.e., having a Bochner density) and the subspace \(\mathcal Z\) of all operators having~\((*)\). The upshot is that \(T\in \mathcal L\) can be written as \(T=T_1 + T_2\) with a ``small'' part \(T_1\in \mathcal R\) and a ``large'' part \(T_2\in \mathcal Z\). (That \(\mathcal R\) is a complemented subspace of \(\mathcal L\) has previously been obtained in a different manner by \textit{Z.~Liu} [J. Oper. Theory 40, No. 1, 3--34 (1998; Zbl 0999.46015)].) The authors hope that their decomposition sheds some light on the long-standing open conjecture that an infinite-dimensional complemented subspace of \(L_1\) is either isomorphic to \(\ell_1\) (in which case it is ``small'') or to \(L_1\) (in which case it is ``large''). The paper would have benefited from proper copy-editing.