Numerical Considerations in Computing Invariant Subspaces
Jack J. Dongarra, Sven J. Hammarling, James Hardy Wilkinson · SIAM Journal on Matrix Analysis and Applications · 1992
This paper describes two methods for computing the invariant subspace of a matrix. The first method involves using transformations to interchange the eigenvalues. The matrix is assumed to be in Schur form and transformations are applied to interchange neighboring blocks. The blocks can be either one by one or two by two. The second method involves the construction of an invariant subspace by a direct computation of the vectors, rather than by applying transformations to move the desired eigenvalues to the top of the matrix.