Fast Solvers for Models of Thermosolutal Convection*

Numerical simulation provides insight into the effect physical parameters have on fluid flows under conditions that often make physical experiments and theory intractable. However, these simulations are computationally demanding and in order to extend their applicability, highly scalable and efficient numerical methods are being developed. We discuss a novel block preconditioner based on domain decomposition and fast diagonalization that can be used to accelerate iterative solution methods. We then demonstrate how this technique provides an efficient means of simulating steady fluid flows, and discuss how this can be used in solving models for thermosolutal convection.