Discrepancy Between API and Website for Stable Structures (e.g., Ag Metal)

Hi Materials Project team,

I’m encountering a discrepancy between the results returned via the API and what is shown on the website when querying for the most stable structure of a pure element.

Using the MPRester API, I ran the following query to retrieve structures with:

  • A single element (e.g., Ag)
  • energy_above_hull = 0.0
  • formation_energy = 0.0

Here is the code snippet I used:

from mp_api.client import MPRester

elements = ["Ag"]

with MPRester() as mpr:
    structure_dict = {}
    structure_id_dict = {}
    for element in elements:
        docs = mpr.materials.summary.search(
            chemsys=element,
            num_elements=1,
            elements=[element],
            energy_above_hull=(0.0000, 0.0000),
            formation_energy=(0.0000, 0.0000),
            fields=["energy_per_atom", "e_total", "material_id"]
        )
        for doc in docs:
            print(doc.material_id, doc.energy_per_atom)

When I run this for Ag, the result includes mp-8566, but I noticed that on the website, the most stable silver structure appears to be mp-989737, which also has an energy above hull of 0 eV.

This leads me to a few questions:

  1. Why does the API return mp-8566 as the stable structure for Ag, while the website seems to prefer mp-989737?

  2. For selecting a representative or most stable structure for elemental Ag, which material ID should I use in my analysis?

Any guidance or clarification would be greatly appreciated!

Best regards,
Guangsheng

Hi @Guangsheng_Liu,

Which part of the website is showing mp-989737 as the most stable entry? The materials explorer shows mp-8566 at 0 ev for the energy above hull with mp-989737 being slightly higher

Are you viewing the phase diagram app? Changing the Functional dropdown will change which hull you are viewing.

I do see that mp-989737 has the lowest formation energy for the GGA/GGA+U hull, while mp-8566 has the lowest formation energy for the GGA/GGA+U/r2SCAN mixing scheme and r2SCAN hulls.

Depending on your analysis it may make sense to limit your queries to one functional type.

Thank you for the clarification!

I realized that when I searched for Materials Project using Microsoft Bing, it directed me to the legacy version of the Materials Project website. In that version, mp-989737 is listed as the most stable structure for Ag.

This explains the discrepancy I noticed between the API results and what I saw on the website. I now understand that the API reflects the current version of the database, whereas the legacy site may show outdated or deprecated entries.

Thanks again for your help and for maintaining such a valuable resource!


Ah, makes sense. Happy to help.