testing PAW against FP-LAPW
Posted: Sat Mar 05, 2011 11:29 am
Hi,
In the new tutorial PAW3, it's reported that, in the comparison with FP-LAPW elk code, one must take care of (among other things):
- Match the elk muffin-tin radii and the PAW cutoff radii.
- freeze the core in the elk.
I have a question about this:
The PAW is based on the frozen core approximation which proved to be able to reach the FP-LAPW accuracy when the atomic data sets are conveniently constructed while keeping the efficiency of the pseudopotential methods. In the case of agreement of the PAW and the FP-LAPW this shows the very good quality of the frozen core approximation as the FP-LAPW allows the core relaxation. In the case of disagreement, this means that either the PAW datasets is not of good quality or that the frozen core approximation is no longer valid which sets the limit for PAW. Now, if we freeze the core in the FP-LAPW method we would no longer consider it as a benchmark as it becomes simply an other frozen core approximation method without any advantage compared to PAW.
thank you for corrections and comments.
In the new tutorial PAW3, it's reported that, in the comparison with FP-LAPW elk code, one must take care of (among other things):
- Match the elk muffin-tin radii and the PAW cutoff radii.
- freeze the core in the elk.
I have a question about this:
The PAW is based on the frozen core approximation which proved to be able to reach the FP-LAPW accuracy when the atomic data sets are conveniently constructed while keeping the efficiency of the pseudopotential methods. In the case of agreement of the PAW and the FP-LAPW this shows the very good quality of the frozen core approximation as the FP-LAPW allows the core relaxation. In the case of disagreement, this means that either the PAW datasets is not of good quality or that the frozen core approximation is no longer valid which sets the limit for PAW. Now, if we freeze the core in the FP-LAPW method we would no longer consider it as a benchmark as it becomes simply an other frozen core approximation method without any advantage compared to PAW.
thank you for corrections and comments.