Suggested cutoff for high level VASP calculations

Question

I am new at VASP, and apparently, cutoff convergence is not something like kpoint convergence. Is using ENCUT=1.5xENMAX can be justified? For RPA, GW and MP2 calculations, I cannot use larger values.

Answer 1

Typically, the largest you would need for convergence, is ENCUT = 1.3 * ENMAX. This is the same as specifying as the PREC = ACCURATE. Going beyond these typically only changes the energy of the system of very few meV.

Answer 2

I would go farther and ask, why is cutoff convergence unlike kpoint convergence? You can very easily do a convergence study based on energy cutoffs. Infact, this is often how ENMAX is determined in a code.

Here is an example from the GPAW documentation of the convergence of the Ni PAW potential. As you can see, the error drops as the energy cutoff increases, you should determine what an acceptable error for your work.

You can also look at the convergence of things such as band structures relative to ENCUT. In general though, as long as the ENCUT is consistent, things should at least be comparable.