Wrong density profile from the default pyr-ringwood88.txt

[Minerallo]

Hi everyone,

with Poulami Roy @poulamiiroy, Bernhard Steinberger and Mauele Faccenda we found a possible problem with the default lookup table pyr-ringwood88.txt used in data/material-model/steinberger, which can be accessed e.g. via the Steinberger cookbook. The provided density profile shows phase transitions from the mantle transition zone at the wrong depths (given a present-day adiabatic temperature profile) compared to the PREM profile (e.g. 410 is at 500km depth or almost non-existent, figure below provided by Bernhard Steinberger), as well as higher densities in the lower mantle. How should we proceed? I could ask Bernhard to share the new profile if necessary and eventually it will be good to have a file with the history of how these profiles are computed to understand where the errors come from and why we get 2 different profiles supposedly from the same compositional data.

cheers,
Michaël

[gassmoeller]

Hi @Minerallo, thanks for reporting this. I think I need a bit more context in order to suggest how we proceed. The file in the data directory you are referencing was generated a long time ago by me using the then recent Perplex version 6.6.6. It is of course possible that something changed in Perplex or that I didnt use it correctly to generate that table. However, we will need more than a new profile, we would need a whole new table. Can you explain how the new purple profile was generated? Was that also run through ASPECT from a new data table? Or generated directly from perplex as a profile?

[Minerallo]

Hi @gassmoeller, Bernhard has recalculated a new corrected lookup table (cc) and the purple profile is resulting from the model run and was extracted via paraview (Bernhard could you confirm and tell us more about how the lookup table was generated ?).
pyr-ringwood88d.txt

[bernhard-steinberger]

I just converted and interpolated the table from the Dannbert et al. (2022 I think, with Carolina and Lars) paper from the entropy-pressure space to temperature-pressure space.

[gassmoeller]

The difference in the tables makes sense to me then. The table that is currently in the repository uses the pyrolite composition published in Ringwood and Irifune 1988, which is different from the one used in Dannberg, 2022 (that one uses the composition used in Xu, 2008, which itself references Workman and Hart, 2005 as source). I did a quick check and the major element composition is quite a bit different (e.g. SiO2 mol% 38.71 in Xu et al. and 44.6 in Ringwood). In addition the mineral database used is likely different, and the software package (Perplex vs Hefesto) is different. So all in all a lot of places that can explain a significant difference in the outcome. That being said, I am convinced that your table is more accurate not just because of the PREM profile, but also because it uses newer data and software. We should replace the table in the repository, but unfortunately it is used in quite a number of places, in particular at least two cookbooks and it is the default material for a material model. In order to replace the table we would need to:

1. Add the new data table, including a correct name (it is no longer appropriate to name it pyr-ringwood88 since it is not based on that paper). We should also try to shrink the table a bit. The current one has 5.5 MB and yours is about the same size. We should try to not add large data files to the repository often, otherwise ASPECT will be a giant download at some point.
2. Change the default data table in the grain size material model and the thermodynamic lookup table equation of state.
3. Update the cookbooks steinberger, tomography_based_plate_motions, and visualizing_phase_diagram and rerun them to update the output pictures.
4. Update all the test results that will change and check if any other benchmarks or cookbooks are affected.

I am happy to help, but in particular step 3 will need some time. Would anyone of you be willing to help?

[poulamiiroy]

@gassmoeller Thank you for the detailed explanation! it all makes sense to me now. I’m happy to help with this, but since my PhD defense is coming up soon, I’ll need some time before I can actively contribute. I can start working on the updates after my defense from mid November.