Additive preservers of non-zero decomposable tensors (Q2463616)

A standard technique in preserver problems is a reduction to maps that preserve the set of rank-one operators. In tensor notation, rank-one operator equals \(x\otimes f\in U\otimes U^\ast\). There are several avenues from here. One can study preservers of higher rank, say as \textit{L. B. Beasley} [Linear Algebra Appl. 107, 161--167 (1988; Zbl 0651.15004)]. In another direction, one can study maps that preserve decomposable tensors \(x_1\otimes\dots\otimes x_m\) on higher order tensor product spaces \(\bigotimes_{i=1}^m U_i\). This was done by \textit{R. Westwick} [Pac. J. Math. 23, 613--620 (1967; Zbl 0178.03502)], who examined linear preservers of decomposability. Unlike assuming linearity in advance, the emphasis in the paper under review is on additive maps. Several structure theorems are proven. The main one is Theorem 2.4, which classifies additive maps \(\Phi:\bigotimes_{i=1}^m U_i\to\bigotimes_{i=1}^n V_i\) that send nonzero decomposable tensors to nonzero decomposable tensors, and obey certain regularity condition. This condition holds automatically if \(\Phi\) is linear and its image is not contained in some \(y_1\otimes \dots\otimes y_{i-1}\otimes V_i\otimes y_{i+1}\otimes\dots\otimes y_n\). The other theorems in the paper study the effect of varying the hypothesis in Theorem 2.4. To give a glimpse, the author replaces regularity condition with preserving decomposability in both directions, and, as expected, gets even stronger conclusions.