Semi-invariants of matrix differential systems (Q1114053)

Some elementary parts of the theory of linear homogeneous differential equations [cf. \textit{E. J. Wilczynski}, ``Projective differential geometry of curves and ruled surfaces'' (Leipzig 1906)] are generalized from the scalar to the matrix case. Capital letters denote \(n\times n\) matrices. Two differential systems of order m and dimension n, m,n\(\geq 2\), \[ (1)\quad Y^{(m)}(x)+\sum^{m}_{k=1}\left( \begin{matrix} m\\ k\end{matrix} \right)P_ k(x)Y^{(m-k)}(x)=0, \] \[ (2)\quad V^{(m)}(x)+\sum^{m}_{k=1}\left( \begin{matrix} m\\ k\end{matrix} \right)Q_ k(x)V^{(m-k)}(x)=0, \] are row-equivalent, in the x-interval I, if there exists a sufficiently smooth matrix S(x), invertible on I, such that \(Y(x)=S(x)V(x).\) A polynomial \(F(P_ 1,...,P_ m,P_ 1',...,P_ m',...)\) with scalar coefficients is a matrix semi-invariant (m.s.i.), if \(S_ p^{-1}FS_ p\) is invariant, up to a similarity given by a constant matrix, in each class of row-equivalent matrices. Here \(S_ p(x)\) is a solution of the equation \(S'(x)=-P_ 1(x)S(x).\) A series of m.s.i., \(P_{k,\ell}(x)\), \(k=2,...,m\), \(\ell =0,1,...\), is given inductively. Every m.s.i. is similar to a polynomial in the matrices \(P_{k,\ell}(x)\). Scalar semi-invariants (s.s.i.) are scalar functions \(f(p_{kij},p'_{kij},...)\), where \(p_{kij}(x)\), \(k=1,...,m\); \(i,j=1,...,n\), are the entries of the matrices \(P_ k(x)\). A sequence of s.s.i. \(a_{k\ell i}(x)\), \(i=1,...,n\) are obtained from the m.s.i. \(P_{k,\ell}(x)\). The s.s.i. are computed in the case \(m=n=2\) (loc. cit. pp. 95-103).