I calculated energies per atom of bulk configurations of transition metals in VASP in order to use them as reference energies for adsorption on surfaces, etc. I downloaded POSCAR files from the material project, calculated their energies in VASP, and divided by the number of atoms in the unit cell. I used PAW PBE pseudopotentials. This is what I got.

enter image description here

Then I decided to do similar calculations in DFTK.jl also in PBE. It uses Hartwigsen-Goedeker-Hutter (HGH) pseudopotentials. DFTK.jl gives energies in Hartree, I multiplied them by 27.2114 and got the following picture with completely different numbers:

enter image description here

I understand that different pseudopotentials use different references and cannot be straightforwardly compared. I just did not expect that the numbers would be that different.

Is there a way to take one pseudopotential, make some calculation on it, and compare it with another pseudopotential or to some calculation on that pseudopotential?

I just want know how to check if I have any mistakes in calculations, and to understand what to do and what not to do when using different software for different parts of calculations. For example, I used the same cutoff energies but it probably does not make sense because the cutoff should be different for different pseudopotentials.

Moreover, for certain atoms, DFTK.jl has different versions of a hgh pseudopotential and each gives very different result, so I am confused, which of them to use., e.g. for Ir:

ir-q9:   -20.225468 Ha/atom
ir-q17: -104.765948 Ha/atom 
  • $\begingroup$ As the energy is a potential energy, I think that don't make sense to compare one each other. It could be preferable to compare a property you calculate with each pseudos with the experimental value (band gap, lattice parameters, etc.). $\endgroup$
    – Camps
    Jul 2, 2022 at 14:38
  • 2
    $\begingroup$ Calculate the energy of all atoms in your system, separately, using both pseudopotentials. Subtract off the energy of all lone atoms from each bulk energy calculation using the appropriate pseudopotential. The result is the cohesive energy, which can be compared across different pseudopotentials or model chemistries. This can also be compared with experiment if such data is available. $\endgroup$
    – Hayden S
    Jul 2, 2022 at 15:34

1 Answer 1


As a first approximation, the total energy will depend on the number of electrons in valence for a given pseudo -potential. This should explain the huge difference in between the two set of calculations.

Then, even if you consider pseudo potentials with the same number of electrons in valence there are many details which will influence the total energy. What makes more sense is to compute energy differences which can be related to some physical quantities.

As suggested in one of the comments, you can for example compute the cohesive energy, i.e. The energy gain to form the material, by subtracting the energy of free atoms to the energy of the same atoms forming the solid. This will be comparable in between different pseudo potentials.


You must log in to answer this question.

Not the answer you're looking for? Browse other questions tagged .