Article ID: | iaor20119939 |
Volume: | 50 |
Issue: | 2 |
Start Page Number: | 193 |
End Page Number: | 221 |
Publication Date: | Oct 2011 |
Journal: | Computational Optimization and Applications |
Authors: | Kanzow Christian, Buchholzer Hannes, Knabner Peter, Krutle Serge |
Keywords: | engineering, programming: nonlinear, matrices, simulation: applications |
The semismooth Newton method was introduced in a paper by Qi and Sun (1993) and the subsequent work by Qi (1993). This method became the basis of many solvers for certain classes of nonlinear systems of equations defined by a nonsmooth mapping. Here we consider a particular system of equations that arises from the discretization of a reactive transport model in the subsurface including mineral precipitation‐dissolution reactions. The model is highly complicated and uses a coupling of PDEs, ODEs, and algebraic equations, together with some complementarity conditions arising from the equilibrium conditions of the minerals. The aim is to show that this system, though quite complicated, usually satisfies the convergence criteria for the semismooth Newton method, and can therefore be solved by a locally quadratically convergent method. This gives a theoretical sound approach for the solution of this kind of applications, whereas the geoscientist’s community most frequently applies algorithms involving some kind of trial‐and‐error strategies.