Some new algorithms for the spectral dichotomy methods (Q1855438)

The authors present new algoritms for the spectral dichotomy methods. Given a square regular matrix \(A\) and a positively oriented contour \(\gamma\) in the complex plane, the spectral dichotomy methods applied to \(A\) and \(\gamma\) consist in determining whether \(A\) possesses eigenvalues on or in a neighborhood of \(\gamma\). In section 2 they give a detailed description of the spectral dichotomy of a matrix with respect to a circle. The results of this section are used througout the paper. In section 3 they discuss the spectral dichotomy of a matrix \(A\) with respect to the imaginary axis. In section 4 they develop a new computational tool to compute the deflating subspaces of a regular matrix pencil of the form \(\lambda B-A\) corresponding to the eigenvalues inside and outside the unit disk. In section 5 they are interested in computing the projector onto the deflating subspace of \(\lambda B-A\) associated to the eigenvalue inside the ellipse. Each section terminates with an algorithms easily implementable that determines the projector associated with the eigenvalues enclosed by (excluded from) the region of interest and the dichotomy condition number that describes the numerical quality of the computed projector.