I recently found from the density output of my visco_plastic model that the density seems to be calculated based on the T of the cell rather than the T of a specific point in the cell.
As you can see in the attached snapshot, this is the output of the initial timestep, which shows a linear increase in T from 273 K at the top to 1573 K at 10 km depth. Correspondingly, its density should also increase linearly from the top to 10 km depth. However, the output density is increasing in stages (the green arrow). (This also happens for the viscosity output).
In the source code (out.densities in multicomponent_incomressible.cc), it does use the T value of the point to calculate the density. But I don’t understand why the density coming from the output is not? I guess this may have something to do with the low resolution of the model, because after I tried to increase the resolution it did appear to have more steps, but it was still not the same as the linear growth of T.
I would be very grateful if someone could give me some explanation? Thanks!
I forgot to mention something more interesting: the model here uses the latest version of ASPECT and dealII 9.3.1. However, this problem does not exist in the same model that uses an older version of ASPECT (commit-04326c1) and dealII 9.2.0.
@gassmoeller Rene, The pressure scaling calculation option you updated (#3681 in Github) doesn’t seem to work for my model either, but that’s another issue.
Can you post an input file that reproduces the issue?
And just to make sure: I assume this is without any cell-wise averaging of material properties? So ASPECT averages the material properties even though it is not supposed to?
If you set Material averaging = harmonic average, ASPECT will cell-wise average the material properties, leading to the behavior you describe above. If you do not want that, just remove the line (or if you are using the GMG solver, which needs averaging of the viscosity, change it to harmonic average only viscosity).
