On sesquilinear forms over fields with involution (Q2463617)

The following result was proven a couple of times in the literature, see in particular \textit{R. Gow} [Linear Multilinear Algebra 8, No. 4, 329--336 (1980; Zbl 0432.15004)] and \textit{D. Ž. Đoković, K.D. Ikramov}, [J. Algebra 257, 97--105 (2002; Zbl 1016.15028)]: Every square matrix A over a field k is congruent to its transpose, i.e., there exists a non-singular matrix \(U\) such that \(U^TAU=A^T\). Moreover \(U\) can be chosen such that \(U^2\) is the identity. This was extended to fields with involution by \textit{R. A. Horn, V. Sergeichuk} [Linear Algebra Appl. 389, 347--353 (2004; Zbl 1068.15007)]: Let \(k\) be a field of characteristic \(\neq 2\) with an involution \(\sigma\), which we let act on matrices entry-wise. Write \(U^\ast=(U^T)^\sigma\) where \(U\in M_n(k)\). Then given \(A\in M_n(k)\) there exists \(U\in \mathop{GL}_n(k)\) such that \(U^\ast AU=A^T\) (we say that \(A\) and \(A^T\) are \(\ast\)-conjugate) and \(U\) can be chosen such that \(UU^\sigma=I\). In the paper under review, a new proof of the above result is given. It relies on a classification of so called non-split matrices, by which we mean those matrices, which are not \(\ast\)-conjugate to proper direct sums of two matrices. This classification forms the main part of the paper.