Semilocal and global convergence of the Newton-HSS method for systems of nonlinear equations. (Q2889387)

The paper is devoted to the Newton-HSS method for solving systems of nonlinear equations \(F(x)=0\) with large sparse non-Hermitian positive definite Jacobian matrix. The method is described in the paper by \textit{Z. Bai} and \textit{X. Guo} [J. Comput. Math. 28, No. 2, 235--260 (2010; Zbl 1224.65133)] where only the local convergence is proved. In this paper the authors prove semilocal and global convergence under reasonable assumptions on function \(F\). First, it is proved that the iteration sequence generated by the Newton-HSS algorithm is well-defined and converges to the solution of the system of nonlinear equations. The initial point needs to satisfy some conditions. Then the authors give the Newton-HSS method with a backtracking strategy utilizing a sufficient decrease condition on the norm of \(F\). This algorithm is based on that in proposed by \textit{S. C. Eisenstat} and \textit{H. F. Walker} [SIAM J. Sci. Comput. 17, No. 1, 16--32 (1996; Zbl 0845.65021)], uses several types of the so-called forcing terms, and has the property that the iteration converges to a root of \(F\) for any initial point.