some more fixes

benchmarks/shear_bands/magmatic_shear_bands.prm

@@ -152,10 +152,6 @@ subsection Model settings
   set Include melt transport                  = true
 end
 
-subsection Melt settings
-  set Melt transport threshold                = 1e-4
-end
-
 subsection Postprocess
 
   set List of postprocessors = visualization,composition statistics,velocity statistics, shear bands growth rate

benchmarks/shear_bands/shear_bands.prm

@@ -137,10 +137,6 @@ subsection Model settings
   set Include melt transport                  = true
 end
 
-subsection Melt settings
-  set Melt transport threshold                = 1e-4
-end
-
 subsection Postprocess
 
   set List of postprocessors = visualization,composition statistics,velocity statistics,shear bands statistics

benchmarks/solitary_wave/solitary_wave.prm

@@ -148,10 +148,6 @@ subsection Model settings
   set Include melt transport                  = true
 end
 
-subsection Melt settings
-  set Melt transport threshold                = 1e-4
-end
-
 
 subsection Postprocess
 

cookbooks/global_melt.prm

@@ -60,22 +60,6 @@ end
 
 ##################### Settings for melt transport ########################
 
-# In models with melt transport, the are additional 'Melt settings' that 
-# let us specify parameters related to the melt transport computations. 
-# One of these is the 'Melt transport threshold': If the porosity is lower
-# than the value specified here, melt will no longer move with a different
-# velocity than the solid, and the system of equations will revert to the
-# usual Stokes equations. Only for porosity larger then the threshold, 
-# the two-phase flow equations apply. 
-# We set the threshold to zero, so the two-phase flow equations are solved 
-# everywhere in the domain. For the limit of the porosity going to zero, 
-# these equations revert to the Stokes equations, and setting the limit to
-# zero has the advantage that there is no sharp boundary between the 
-# region where melt is present and the regions where no melt is present.
-subsection Melt settings
-  set Melt transport threshold                = 0.0
-end
-
 # In models with melt transport, we always need a compositional field with
 # the name 'porosity'. Only the field with that name will be advected with
 # the melt velocity, all other compositional fields will continue to work

doc/manual/cookbooks/global_melt/global_melt.prm

@@ -40,17 +40,6 @@ end
 
 ##################### Settings for melt transport ########################
 
-# In models with melt transport, the are additional 'Melt settings' that 
-# let us specify parameters related to the melt transport computations. 
-# One of these is the 'Melt transport threshold': If the porosity is lower
-# than the value specified here, melt will no longer move with a different
-# velocity than the solid, and the system of equations will revert to the
-# usual Stokes equations. Only for porosity larger then the threshold, 
-# the two-phase flow equations apply. 
-subsection Melt settings
-  set Melt transport threshold                = 0.0
-end
-
 # In models with melt transport, we always need a compositional field with
 # the name 'porosity'. Only the field with that name will be advected with
 # the melt velocity, all other compositional fields will continue to work

doc/modules/changes/20180406_jdannberg

@@ -0,0 +1,17 @@
+Overhaul of the melt solver. 
+Changed: Modelwith melt transport now have a new preconditioner, 
+and the linear system is solved in a different way: Based on the 
+Darcy coefficient, each cell is classified as a melt cell 
+(where the melt transport equations are solved) or not a melt 
+cell (in this case, the compaction pressure dofs are constrained
+to zero, and the equations that are solved are the Stokes system
+without any melt-related terms). To achieve better convergence 
+for low melt fractions, the compaction pressure p_c is replaced
+by a scaled compaction pressure p_c_bar, so that the linear 
+system we are solving is different than before, and the solution 
+vector now contains p_c_bar instead of p_c. 
+The advantages of these changes are much lower iteration counts in 
+models with low or zero porosity, and that we are now solving 
+the Stokes system if no melt is present. 
+<br>
+(Juliane Dannberg, 2018/04/06)

source/simulator/melt.cc

@@ -1573,7 +1573,7 @@ namespace aspect
     // now subtract the ones that are nonzero as computed above:
     for_constraints.subtract_set(nonzero_pc_dofs);
 
-    // and constrain those.
+    // and constrain the remaining dofs (that are not in melt cells).
     constraints.add_lines(for_constraints);
     constraints.close();
   }