Is the higher the ENCUT, the higher the accuracy we get?

[KurtG]

Normally, due to ENCUT controls the size of plane wave basis sets, the answer for this question is yes.

However, in Dr. Georg Kresse's slides for the VASP course (Session 4 and Session 11), he mentioned that the pseudopotentials are always optimized for the energy cutoff 'using a minimal basis set'. And 'the PAW potentials are optimized to work at a cutoff of 250-300 eV'. So I doubt whether a result with very high cutoff energy is meaningful.

For example, I calculated some rare earth oxides, especially the ones with extended 4d and f electrons. I find the ENCUT convergence is acceptable (<20 meV for a 20-atom cell) only when the ENCUT is very large. For instance, the total energy of SrRuO3 converges at ENCUT~850 eV! 500 eV seems to converge well, but if further increase the cutoff we will get ~200 meV total energy gained. Please see the figure here.

So, in this case, whether the result with a higher cutoff has a better accuracy? or worse?

Thank you.

[alex]

Hi KurtG,

some remarks:
a) we prefer to look at differences of total energies in our trade rather than absolute values
b) differencies are mostly well converged at Georg's typical cut-offs
c) by the shape of your figure you probably missed the so-called 'kinetic energy error' written in the OUTCAR and added to the total energy (to account for a non-complete basis set)

Cheers,

alex

[KurtG]

(a & b)yes, you are right. Most of the time only the differencies make sense.

(c) Even take into account the kinetic energy error contribution, there is still 30 meV difference between ENCUT=500 and 1200.

Thank you for your reply.
[quote="'smallblacktext'>[ Edited Fri Jan 06 2012, 04:11PM "]</span>