Changes between Version 313 and Version 314 of doc/app/initialization_parameters

Timestamp:

Mar 20, 2017 10:19:26 AM (8 years ago)

Author:

hellstea

Comment:

--

doc/app/initialization_parameters

@@ -1350 +1350 @@
 }}}
 {{{#!td style="vertical-align:top"
-'dirichlet'
+'dirichlet'\\\\
+('neumann' for nest domains)
 }}}
 {{{#!td
 Top boundary condition of the perturbation pressure.\\\\
 Allowed values are '' 'dirichlet' '' (p(k=nz+1)= 0.0) or '' 'neumann' '' (p(k=nz+1)=p(k=nz)).\\\\
-Simultaneous use of Neumann boundary conditions both at the top and bottom boundary ([#bc_p_b bc_p_b]) yields no consistent solution for the perturbation pressure in case that the multigrid method is used for solving the Poisson equation (see [#psolver psolver]), and should be avoided. Since at the bottom boundary the Neumann condition  is a good choice (see [#bc_p_b bc_p_b]), in that case, a Dirichlet condition should be set at the top boundary.
+Simultaneous use of Neumann boundary conditions both at the top and bottom boundary ([#bc_p_b bc_p_b]) yields no consistent solution for the perturbation pressure in case that the multigrid method is used for solving the Poisson equation (see [#psolver psolver]), and should be avoided. Since at the bottom boundary the Neumann condition  is a good choice (see [#bc_p_b bc_p_b]), in that case, a Dirichlet condition should be set at the top boundary.\\\\
+\\\\
+In case of nested run, the default is not 'dirichlet', but 'neumann' instead.\\\\
+Note that therefore the Poisson equation for perturbation pressure has all-Neumann boundary condition in nest-domains. This means that the solution is only determined up to a constant. Or in other words, the average value of the pressure is arbitrary. This implies that the convergence rate is usually somewhat lower than when Dirichlet condition is used at least on one of the boundaries. Users should pay attention to this and adjust the steering parameters of the multigrid solver carefully to achieve sufficient convergence without sacrificing too much computing time for the perturbation-pressure solution. 
 }}}
 |----------------