10.3 What kind of ``technical'' errors do exist, overview

Technical errors fall into four categories

• Errors due to k-points sampling. This will be discussed in section 10.6. Mind that the errors due to the k-points mesh are not transferable i.e. a k-points grid leads to a completely different error for fcc, bcc and sc. It is therefore absolutely essential to be very careful with respect to the k-points sampling.
• Errors due to the cut-off ENCUT. This error is highly transferable, i.e. the default cutoff ENCUT (read from the POTCAR file) is in most cases save, and one can expect that energy differences will be accurate within a few meV (see section 10.4). An exception is the stress tensor which converges notoriously slow with respect to the size of the plane wave basis set (see section 9.6).
• Wrap around errors (see section algo-wrap). These errors are due to an insufficient FFT mesh and they are not as well behaved as the errors due to the energy cutoff (see section 10.4). But once again, if one uses the default cutoff (read from the POTCAR file) the wrap around errors are usually very small (a few meV per atom) even if the FFT mesh is not sufficient. The reason is that the default cutoffs in VASP are rather large, and therefore the charge density and the potentials contain only small components in the region where the wrap around error occurs.
• Errors due to the real space projection. Real space projection always introduces additional (small errors). These errors are also quite well behaved i.e. if one uses the same real space projection operators all the time, the errors are almost constants. Anyway, one should try to avoid the evaluation of energy differences between calculations with LREAL=.FALSE. and LREAL=On/.TRUE (see section 7.34). Mind that for LREAL=On (the recommended setting) the real space operators are optimized by VASP according to ENCUT and PREC and ROPT i.e. one gets different real space projection operators if ENCUT or PREC are changed (see section 7.34).

In conclusion, to minimize errors one should use the same setting for ENCUT, ENAUG, PREC, LREAL and ROPT throughout all calculations, and these flags should be specified explicitly in the INCAR file. In addition it is also preferable to use the same supercell for all calculations whenever possible.