The PHAST Solution Method dialog box is used to control which solver is used and to set the parameters for the solver. The PHAST Solution Method dialog box is displayed by selecting Model|PHAST Solution Method…
The user can select either the direct or iterative solver. For large problems, the iterative solver is more efficient.
The user can choose to include cross dispersion terms by checking the appropriate checkbox. For problems where there is flow at an angle to the grid, including the cross-dispersion terms can lead to a more accurate solution at the cost of an increased execution time. However, the cross-dispersion terms can also lead to negative concentrations. These will be set to zero in the chemical calculations, which leads to global mass-balance errors.
Edit boxes are available to set the space-differencing and time-differencing weighting. Space-differencing weighting and time-differencing weighting are important parameters because they help control the oscillation and numerical dispersion in the numerical solution. In general, centered-in-space and centered-in-time weightings (when both parameters are 0.5) reduces numerical dispersion, but can lead to oscillation of the solution. Upstream-in-space (space differencing = 0) and backwards-in-time (time differencing = 1) do not produce oscillations, but have more numerical dispersion (Parkhurst and others, 2004).
If the iterative solver is used, a tolerance must be specified in the related edit box. The tolerance determines when the solver has converged to a solution to the difference equations. The tolerance must be sufficiently small to obtain an accurate solution.
If the iterative solver is used, the number of search directions to be saved (Save Directions) between restarts of the iterative solver must be specified in the Save directions edit box.
Sometimes iterative solvers cannot converge for a particular problem. The user specifies the maximum number of iterations that will be allowed before the solver quits in the Maximum iterations edit box.
The Rebalance fraction option applies only to the parallel version of PHAST. After each time step, PHAST evaluates how to optimally rebalance the cells among the processors for the chemical calculation. It estimates a number of cells to transfer from one processor to the next. The Rebalance fraction is multiplied times the estimated number of cells to obtain the actual number of cells transferred. The default value is 0.5.
Rebalance by cell is used optionally used for the parallel version of PHAST. By default, PHAST looks at the total time taken to run chemistry on a block of cells and from that calculates the average time per cell, which is used to rebalance the load among the available processors. The new algorithm measures the time for each cell individually and rebalances based on this information. The new method should only be needed for kinetic problems where some cells take much longer to run than other cells. If Rebalance by cell is checked the new algorithm is used for parallel runs. Otherwise the original algorithm is used to rebalance workload for parallel runs. The default value is unchecked.
