Two new variants of nonlinear inexact Uzawa algorithms for saddle-point problems (Q1864503)

The authors present a study of nonlinear inexact Uzawa methods for iterative solving linear saddle-point problems. An essential improvement on the convergence results of \textit{J. Bramble, J. Pasciak} and \textit{A. Vassilev} [SIAM J. Numer. Anal. 34, 1072-1092 (1997; Zbl 0873.65031)] for a known nonlinear inexact Uzawa algorithm is presented. The authors present two new algorithms which are a combination of the known nonlinear inexact Uzawa method and the conjugate gradient method. Both algorithms not require any apriori estimates on the minimal and maximal eigenvalues of the preconditioned system, including the preconditioned Schur complement. Efficiency of the new algorithms are illustrated on two numerical tests: two-dimensional Stokes problem and a purely linear system of algebraic equations.