Dear Afshin,

  Did you have a look in the PSLibrary, http://www.quantum-espresso.org/pseudopotentials/pslibrary/ ?

    Greetings from Paris,

       apsi

Dear Ari,

Thank you very much. But my inquiry was, " how to select an appropriate type of pseudo-potential between all the possible ones, to calculate defect properties", e.g. in PSlibrary we can use either ultrasoft, paw, coloumbic, ... type PS, but which of them will produce better results. I wonder is there any specific standard?

Greetings from Tehran,

Afshin 

Hi Afshin,

The best pseudopotential is the one that best reproduces known properties of the system of interest.

- Adib

Dear Afshin,

  Like Adib said, it depends. In general I would expect the accuracy to be the best with a PAW data set/method, then "Vanderbilt"/USPP pseudo potential, of course the -Z/r is the correct potential but it cannot be used in practical systems, only in some tests with the very light elements (I think I have gone up to oxygen). Norm-conserving pseudo potentials are good for some elements, sometimes they are required, when a piece of code has not been generalised to handle PAW/USPP (they usually come together with several new terms in the analytical expressions, thus also in the implementation). Relative speed, well depends on the material (elements needed there; semi-core states, ...); in the light and easy elements one migth pay some penatly when using PAW and USPP relative to norm-conserving pp, in the more complicated elements the cut-off energy for the wave functions can bring the computational cost down - and one has better accuracy to be expected thanks to more degrees of freedom (and/thus more reference energies, in PAW the radial grid, ...).

  There is no general answer, as you see above; it needs some testing and effort. Even if those are technical issues, by studying them one also understands better the underlying assumptions and even the physics/chemistry.

    Greetings,

       apsi