I am testing some stuff with the ‘Latent heat’ material model. I have found some curious stuff that escapes a bit my understanding on what ASPECT is doing.
I have implemented a layered system with one single phase-change interface. My model is not compressible at least in the definition of the parameter file, no adiabatic heating, etc.
Upon plotting results in Paraview, carefully inspecting (plotting in different ways, and using ‘plot over line’) I realized that the density is increasing with depth (not solely jumping at the interface, but continuously inside the layers). Checking the manual, I realized that there is a default non-zero compressibility(~ +5e-12) I was unaware of.
I think there is nothing wrong with this in principle (physically), but I was trying to isolate everything and my maths did not give me my expected answer. I guess this compressibility is unavoidable in order to allow the software to perform the density jump across the phase change..
Can this compressibility set equal to 0 and have the system run correctly? I am trying to save some SUs on my account, and some explanations would be helpful.
In the latent heat material model there are parameters that control the width of the phase transition (e.g., Phase transition pressure widths or Phase transition widths).
Can you let us know what you assigned for these values and also the other parameters related to the phase transitions?
As I explained, although for my model I did not specify any compressibility nor adiabatic heating, the model is compressible (by default of Aspect settings), something I did not aimed for.
I have implemented 4 models of this kind. In all of them I set the same
Clapeyron slope(-2.6 MPa/K) and the same density jump (350 kg/m3) (it’s
like at the 660 km depth). In 2 of them I define the transition by pressure (2.904e10 Pa), with non-zero pressure widths (2.2e9 Pa); and in
the 2 last ones by depth (660e3 m), with 0 width.
In all
these models, the density jumps at the phase transition (by ~350, as expected), but it varies continuously inside the layers, i guess obviously in response to the default compressibility (5e-12).
Thanks for the follow-up explanation. Just so we are on the same page, you are seeing density variations that would not be caused by temperature alone within each layer (e.g., assuming the thermal expansivity is not zero)?
If you are seeing density variations within the layers not explained by temperature variations alone, it is likely the default compressibility. Setting that value to zero will indeed turn off compressibility (e.g., material model returns that it is an incompressible system).
I suggest running a few low-resolution 2D runs (see test suite) to make sure the chosen parameters are giving exactly what you expect.
Please let us know if something is still amiss after running the tests.