Upper triangularization of matrices by permutations and lower triangular similarity transformations

Let A be an arbitary (square) matrix. As is well known, there exists an invertible matrix S such that S-1AS is upper triangular. The present paper is concerned with the observation that S can always be chosen in the form S=∏L, where ∏ is a permutation matrix and L is lower triangular. Assuming that the eigenvalues of A are given, the matrices ∏, L, and U=L-1∏-1A∏L are constructed in an explicit way. The construction gives insight into the freedom one has in choosing the permutation matrix ∏. Two cases where ∏ can be chosen to be the identity matrix are discussed (A diagonable, A a low order Toeplitz matrix). There is a connection with systems theory.