Numerical Solution of Matrix Polynomial Equations by Newton's Method

From MaRDI portal





The paper presents an algorithm, corresponding to the general Newton- Kantorovich method, for the numerical solution of matrix polynomial equations of the form \(P(X)=\sum^{n}_{r=1}A_ rX^ r=B\) in the unknown \(m\times m\) matrix X, with \(A_ 1,...,A_ n\), B \(m\times m\) complex matrices. In the case of simple solutions (solvents), i.e. where the Fréchet derivative of P(X) is regular, this method converges quadratically in a neighbourhood of any solvent. Each step of the iterative method requires the solution of a linear equation of the form \(\sum^{n}_{r=1}\hat A_ rH\hat B_ r=C\) in the unknown matrix H. Such equations are solved by a transformation into an \(m^ 2\times m^ 2\) linear system of type \(Gx=b.\) Numerical examples and the numerical behaviour of the method are discussed. The proof of the convergence of the method is based on explicit estimates for the Fréchet derivative; these bounds also yield a new existence result for the equation \(P(X)=B\), proved by means of a modified algorithm which involves only the derivative of P at a fixed point. Finally, the method is applied to scalar matrix equations \((A_ r=a_ rI)\), to quadratic equations, and to the problem of finding the square root of B; in comparison with known methods, it is shown that the proposed method is self-correcting.




Cited in
(29)








This page was built for publication: Numerical Solution of Matrix Polynomial Equations by Newton's Method

Report a bug (only for logged in users!)Click here to report a bug for this page (MaRDI item Q3768267)