ABINIT Discussion Forums

Hello,
I'm a beginner in the use of Abinit for quantum chemistry calculations. I wanted to start with the "simple" case of the MgO structure (Fm3m). I'm performing the calculations in the PAW mode, with the GGA-made pseudopotentials given on the website. Strange thing is that, these pseudopotentials are made to be used (or have been tested) in the case of a MgO (great for me so!), but with performing the optimization of the cell, it's giving me an error message about the overlap of densities of the order of 27% (overlap usually accepted is less than 5%). Is that normal?
In a second time if I'm launching the calculations, putting an overlap acceptance of more than 27% (which seems to be dangerous, because of possible inaccuracy of the results), I'm finding a cell size (acell) of 2.977 Bohr (instead of around 4.2, like used at the beginning).

Are there any corrections to give to the pseudopotential file ?

I'm also trying to do the same type of calculations for the CaO crystal (same structure, Fm3m), and I have to build, this time, the pseudopotential. Any advices?

4.2 Angstrom (not Bohr) is the conventional (fcc) unit cell dimension for MgO.
2.97A is about right for the primitive (rhombohedral) unit cell.

Hum... I see. I've just made one of the most stupid mistake of my life... This way, the overlap totally disappear, that's the best part.
Remains that the lattice parameter obtained is still too small (around 2.99 Angstr) in regards of the value with which we started (4.211 Angstr).
Thank you very much for this first big step anyway!

Which PAW sets are you using exactly? My own Mg and O PAW sets with PBE give a lattice constant 2% longer than experiment and a bulk modulus of 150 GPa (rather than 165 GPa). This is fairly typical behavior for PBE.

I'm actually using the PAW pseudopotentiels given here :
http://www.wfu.edu/~natalie/papers/pwpaw/periodictable/periodictable.html, with the GGA form.
I've also tried to create my own pseudopotentials with Atompaw. For example for the Mg, I've created a pseudo like that :
-Valence electrons : 3s orbitals
- lmax = 0
- Critical radius rc = 2.67 Bohr (the covalent radius of the atom)
- with a pseudization based on the following processes : "custom RRKJ" and "bessel".
- and two basis set wavefunctions (limited with the same critical radius)

With my input file, I'm having the same problem, meaning that the cell is contracting too much, whatever are the pseudopotentials used.

Edit : Isn't that strange to find a lattice parameter of 2% longer in regards of experimental results, as this calculations are made theoritically at 0K ? (so the cell should contract, not extend)

Thomas wrote:I'm actually using the PAW pseudopotentiels given here :
http://www.wfu.edu/~natalie/papers/pwpaw/periodictable/periodictable.html, with the GGA form.
I've also tried to create my own pseudopotentials with Atompaw. For example for the Mg, I've created a pseudo like that :
-Valence electrons : 3s orbitals
- lmax = 0
- Critical radius rc = 2.67 Bohr (the covalent radius of the atom)
- with a pseudization based on the following processes : "custom RRKJ" and "bessel".
- and two basis set wavefunctions (limited with the same critical radius)

With my input file, I'm having the same problem, meaning that the cell is contracting too much, whatever are the pseudopotentials used.

Edit : Isn't that strange to find a lattice parameter of 2% longer in regards of experimental results, as this calculations are made theoritically at 0K ? (so the cell should contract, not extend)

In the same site, there is the result for MgO using their Mg and O PAW GGA PP.
http://www.wfu.edu/~natalie/papers/pwpa ... A_plot.png