The correction energy of Cl2 used in total energy calculations

I noticed that an adjust energy is applied for gas-to-solid reactions (O2, N2, Cl2, …), as shown in the section of “Default corrections: Gases, liquids, and elements”: https://docs.materialsproject.org/methodology/total-energies/.
However, the data is only available for O2, which is listed in the original paper of Wang (Ref. [5] in the above-mentioned webpage), and I could not find the correction energy of others. 
Could the others share the correction energies for N2, Cl2, …?
Thank you!

Hi @Guo-Dong_Zhao. We refer to these energies as our “correction scheme.” You can find the current values in the file MPCompatibility.yaml in pymatgen.

However, please be aware that we have an updated correction scheme that will become the default on MaterialsProject within the next month or two. It will be thoroughly documented in a forthcoming manuscript. You can already find those correction values listed in MP2020Compatibility.yaml, and you can apply them to ComputedEntry using the MaterialsProject2020Compatibility() class.

I hope this helps!

1 Like

Thank you so much!!!

Hi @rkingsbury,

Is the same correction scheme (MP2020Compatibility) applied to the new R2SCAN computed entries on MP as well?

I have received mixed information from materials project website and mp_api. On MP website, for example YbCl2, MP2020 anion correction (Cl)
-0.61 eV/atom is applied to formation energy, but there is no correction information for r2scan calculation, which is marked as the primary calculation on tasks page. I have used the new R2SCAN INCAR to calculate a couple of compounds and applied MP2020Compatibility.process_entries() (custom data), but I didn’t see any energy correction applied to the computed entry.

If my understanding is correct, the structure page still presents GGA/GGA+U mixing data, and there is no correction scheme applied for the new R2SCAN data if retrieved from mpr.thermo.search(material_ids, thermo_types=['R2SCAN']). Is this correct?

If there is any corrections anion, elemental, gas applied to R2SCAN entries, could you kindly indicate how to programmatically apply the same correction to the custom data? Thanks!

Hi @cyrusyc , there are no corrections applied to R2SCAN calculations. MP2020Compatibility.process_entries is actually designed to guarantee that energy corrections are only applied to GGA/GGA+U calculations with the correct (MP-compatible) settings; any other calculations including R2SCAN will be ignored.

For more information, see our documentation page here: GGA/GGA+U/r2SCAN Mixing - Materials Project Documentation

Thank you Dr. @rkingsbury ! This is very helpful.

I have read your both papers (PRM and npj). Is it safe to say that now on MP (v2022.10.28) the raw energy data (say the Total energy -108.35 eV of mp-2050790) of all binary materials retrieved by calling thermo.search(material_ids, thermo_types=['R2SCAN']) are based on your npj paper rule#2 and all binary compounds close to hull (within 20 meV per atom of PBE(+U) ground states) have all been calculated using R2SCAN without any mixing correction? Or we still need to wait for more extensive R2SCAN coverage?

Thanks,
Yuan

Hi Yuan, I’m not up to date on the current state of R2SCAN database coverage, but in general you should always inspect the specific entries that you download.

When you look at a task document (like the one you linked), you are always looking at the raw calculation output. So the Total Energy you refer to is the uncorrected / raw R2SCAN energy. The same would be true if this were a GGA/GGA+U task.

When you look at the materials document (and/or the data displayed on the materials detail page) OR if you download thermo documents from the thermo endpoint, that’s when you will see energy corrections. For any given ComputedEntry, you can always inspect the energy_adjustments attribute to see which corrections, if any, have been applied. If the R2SCAN mixing scheme was used, you will see a correction in there labeled as such.

In this particular case, if you look at the phase diagram on the materials detail page, you can hover your mouse over YbCl2 and see that it is coming from a GGA calc (mp-865716-GGA). You can also expand the item “How was this formation energy calculated” below the phase diagram to see that a Cl anion correction was applied. If you use thermo.search without the R2SCAN filter, you should get this computed entry.

Finally, if you use the Phase Diagram app on the left hand side, you can actually select which set of calculations you want to use to build the phase diagram. If I try this, it is possible to build the phase diagram with only R2SCAN data.

So clearly we have all the necessary data, but the formation energy on the materials detail page is still coming from the GGA calculations. I’m not sure specifically why that’s the case, but there can be technical reasons where some materials cause problems when we build the whole database forcing us to exclude the R2SCANs.

I hope this helps! I’m not directly involved with the MP team anymore but @ruoxiyang is also a good resource for questions about the status of the database and the data you see on the website.

1 Like

Thank you @rkingsbury! This clarifies things a lot! Sorry for shooting so many annoying questions. I was confused by the raw, calculated and corrected, mixed data pulled down from thermo endpoints when I saw mismatches on different corner of MP website. Especially I am curious (worried) how aligned the MP database is to the r2scan mixing rule proposed in your paper and how comprehensive the MP database is since we are going to extract reaction enthalpies of all chloride compounds available on MP. It will be really helpful to confirm hat we have r2scan calculated energy (without any correction and mixing scheme) for all ground state structures.

I will follow your suggestion to programmatically check all compounds from thermo r2scan endpoint and their associated energy corrections (should all be zeros if they are available from r2scan calculation I guess, based on my current understanding).

Thank you again @rkingsbury and wish you a strong start at Princeton!

Thank you Yuan! And no need to apologize. Integrating these two datasets (r2SCAN and GGA) is quite complicated and I know the MP team welcomes feedback from users like yourself about how to make the data sources more clear.

Thread closed due to inactivity, please open a new thread to address related issues.