When a \((1,1)\)-tensor generates separation of variables of a certain metric (Q6141352)

This paper asks and answers the following question: What are the necessary and sufficient conditions on a \((1,1)\)-tensor field \(K^i_j\) defined on an \(n\)-dimensional manifold \(M\) for the existence of a metric \(g_{ij}\) for \(K\) to generate separation of variables for this metric? At minimum \(K\) must have \(n\) distinct eigenvalues, and there must be a coordinate system in which \(K\) is diagonal. The condition that \(K_{ij}\) has distinct eigenvalues means that the discriminant of the characteristic polynomial is nonzero. It is also known that the existence of a local coordinate system such that the tensor field is diagonal is equivalent to the vanishing of the Haantjes torsion (as described in [\textit{ J. Haantjes}, Indag. Math. 17 Nederl. Akad. Wetensch. Proc. Ser. A 5 (8), 158--162 (1955; Zbl 0068.14903)] ). The main results of this paper are explicit formulas that can be calculated in any coordinate system to determine whether a given tensor field \(K^i_j\) with \(n\) distinct eigenvalues at every point and vanishing Haantjes torsion generates a separation of variables in some metric.