some questions about the pseudopotential

[Roc]

Dear everyone,
   Recently, I try to explore one molecule including Cr element, while I have a question about the pseudopotential of Cr. In atk, the TM pseudopotential defines that the valence electron is 4S²3d⁴, while others not, for example, in VASP, the plane wave pseudopotential defines the valence electron 4s¹3d⁵. So, can anyone give some possible reasons or advice. which is better?

   Thank you!     

[nori]

It's difficult to answer your question assertively　but I feel the pseudopotential of VASP is better than one of ATK at least for Cr.
As you know, The ground state of Cr crystal is anti-ferromagnetic.
In my calculation with ATK, each Cr element in unit-cell has 4 [tex]\mu_{\rm B}[/tex] or -4 [tex]\mu_{\rm B}[/tex] magnetic moment, that is unphysical result.
On the other hand, in the calculation with VASP, each Cr element in unit-cell has 1 [tex]\mu_{\rm B}[/tex] or -1 [tex]\mu_{\rm B}[/tex] magnetic moment, that is reasonable result.

[zh]

For the generation of pseudoptential, besides the choice of electron configuration of element, the cut-off radius is also important. It can not be asses which one is the better if without checking the quality of the generated pseudoptential.

You do the following test calculations to check the quality of pseudoptential:
i) calculate the bond length of Cr2 dimer;
ii) calculate the lattice constants and bulk modulus of bulk Cr;
iii) compare the calculated  results with the experimental values.

[Anders Blom]

A lot of pseudopotentials (pps) were updated when ATK 2.2 was released in 2007, and Cr was one of them. It was not found that there was any reason to change the valence configuration to get a reasonable bulk lattice constant etc.

However, the choice of pp is not unique, and it may very well be that for certain applications another valence would be more suitable. You can easily replace the ATK pp with your own (as long as it is defined in the same way); you don't even need to replace the actual pp file, you can define another pp in your script for a particular element.

One must admit that TM pps are not the most accurate ones. In the future, the intention is therefore to make ATK compatible with the large libraries of ultrasoft and hopefully also PAW pps that exist, to make it possible to take advantage of the large work that has gone into tuning these pps for particular situations.

[nori]

I re-calculated a spin polarization of Cr with ATK 10.8b and obtaind [tex]\pm 0.82  \mu_{\rm B}[/tex] at each Cr, which is very reasonable result.
I don't know why but ATK 10.8b is more accurate than ATK 2008.10

[Anders Blom]

It is our hope that you will find that several things have improved in ATK 10.8