Underestimation of hyperfine constant

[hongyang_ma]

Hi,

I'm a newbie in using VASP. I would like to use VASP to calculate the isotropic hyperfine constant of P doped Si by using the keyword "LHYPERFINE". I successfully got the results, however, the calculated hyperfine value is much smaller than the experimental value (76 MHz vs. 117 MHz). I'm wondering what is the reason that causes this underestimation? Is this due to the psedupotential which ignore the core electron contribution? Are there any possible ways to improve the results?

The input files I used and OUTCAR are attached

Files.rar

. Please have a look. Thanks you for your help.

Best Regards,
Hongyang

[marie-therese.huebsch]

Hi Hongyang,

Thank you for bringing up this topic. Perhaps someone else has experience with this calculation and can help.

From what I gathered, the result of LHYPERFINE is very sensitive to the PAW dataset you are using. Therefore, you may try if the GW POTCAR yields different results. A computationally more expensive route could be to make use of hybrid functionals. The advantage of hybrid functionals is that you can probably set the amount of Fock exchange just right to reproduce the experimental value. From a critical standpoint that would not be fully ab initio then, because you selected a parameter to fit the experiment. Besides these suggestions, you could also benchmark against an all-electron code like Wien2k, but I don't know if they provide the analogous output for the hyperfine constant.

Best regards,
Marie-Therese

[hongyang_ma]

Hi Marie,

Thanks for you reply. I will give it a try.

Best,
Hongyang

[marie-therese.huebsch]

Hi Hongyang,

May I ask what you concluded after your tests?

Marie-Therese