A note on matrix inversion (Q789471)

This paper shows that if M, N, U, V are left modules over a ring R and \(\partial:M+N\to U+V\) is an isomorphism given in matrix form as \(\left[ \begin{matrix} A&B \\ C&D \end{matrix} \right]\) then if \(\partial^{-1}\) has the matrix form \(\left[ \begin{matrix} X&Y \\ Z&W\end{matrix} \right]\) it follows that A and W have isomorphic cokernels and kernels. Analogous statements hold for D and X, B and Y, Z and C. This generalizes an earlier result of \textit{R. C. Thompson} [ibid. 44, 197-201 (1982; Zbl 0483.15011)] for matrices over a principal ideal domain. In that paper the condition on kernels and cokernels was replaced by the equivalent statement that the pair of matrices had the same Smith canonical form.