Comment on Pseudo Potentials

[jmontgomery@flsouthern.edu]

This may be a topic that is discussed in more detail later, but I am interested in knowing which pseudo potential to use for a certain application. For example, for the Si example we did for week 1 of DFT, the pseudo potential was Si.pbe-n-kjpaw_psl.1.0.0.UPF, which gave an energy of -93.45256277. This potential is a “PAW”. However, after downloading the SSSP library, the Si pseudo potential provided there is Si.pbe-n-rrkjus_psl.1.0.0.UPF, which is a “USPP” potential. The energy calculated with this potential is -22.83911995, which is hugely different! I would imagine the choice of pseudo potential is as varied and important as the choice of functional.

[Ruben Quiroz]

Did you do proper convergence testing for both potentials? Because I also had hugely different results when using a different potential but they went away when I increased the computation time.

[jmontgomery@flsouthern.edu]

Thank you Ruben… It’s a good point. No, I did not do convergence tests with both potentials. But I did do convergence tests for the halite task using a different pseudo potential for Na and Cl (the one provided in the SSSP-precision download), and I still got considerable differences in the total stress. Perhaps I need to revisit.

[Stefaan Cottenier]

Nice discussion — see the feedback webinar for an answer.

[Author]

Posts