Cohesive energy of La

Materials Project Materials Project Data/API
1

Hi,

material_id formula_pretty atomic_number uncorrected_energy_per_atom isolated_atom_energy cohesive_energy_per_atom run_type functional
mp-26 La 57 -29.74478021 -18.0741 -11.67068021 R2SCAN r2SCAN
mp-26 La 57 -4.936007105 -0.845547 -4.090460105 GGA PBE

I would like to raise a possible issue regarding the cohesive energy of La obtained using get_cohesive_energy(). For elemental La, the function appears to return an r2SCAN cohesive energy of approximately 11.67 eV/atom in magnitude, which seems unusually large. For comparison, the experimental cohesive energy reported in Kittel’s textbook is about 4.47 eV/atom, while the PBE result (from Materials Project) is approximately 4.09 eV/atom and is much closer to the experimental value. This discrepancy may be related to the isolated-atom reference energy used for the r²SCAN calculation. Since the cohesive energies of La-containing compounds are calculated using the La isolated_atom_energy as a reference, a problematic reference value could affect the reported cohesive energies of all the La-related materials. Could the Materials Project team please confirm whether the r2SCAN isolated-atom energy for La has been validated, or whether there are any known issues associated with this value?

Thank you.

2

Hi @wstee, my intuition here is this is an issue with the particular pseudopotential used in the older r2SCAN calculations. This is more or less confirmed looking at the MatPES r2SCAN calculation of mp-26 which has a more reasonable -3.96 eV/atom cohesive energy, and which uses updated pseudos

More evidence comes from the SCAN cohesive energy, -12.21 eV/atom from its mp-26 affiliated task. The SCAN calculations in MP use the same older pseudos as the r2SCAN ones, and as r2SCAN is a simplification of SCAN, they often give similar energetics

MP is currently recomputing all of its materials with the updated computational settings used in MatPES, so this should get resolved as time goes on

2 Likes