VASP INCAR parameters

[Redorb]

Hello fellow USPEX users,

I am using USPEX with VASP and I was wondering if anyone had tailored
their VASP INCAR files for their USPEX runs for better accuracy/
efficiency.

The main VASP tags I am intersted in are:

1.) ISMEAR: The VASP manual states that for metals one should use
ISMEAR=1 but ISMEAR > 1 should be avoided for insulators and
semiconductors. This of course is a problem if you don't even know the
structure you are going to be looking at while using USPEX.

2.)ISMEAR = -5 final self-consistent run to get accurate energies.
When I am doing my relaxations of a crystal via VASP I usually follow
this procedure. Relax crystal using ISMEAR=whatever appropriate. Then
do a final self-consistent calculation using ISMEAR = -5 and PREC=
Accurate to get a final total energy of the system that is most
accurate. Has anybody done their USPEX runs in this way or would this
simply be overkill?

3.) ALGO: This flag has 3 setttings that I am intersted in; Normal,
VeryFast and Fast. The Normal Setting does a slower but more accurate
blocked Davidson block iteration for all electronic optimization
steps. Fast does one minimization step using blocked Davidson then
switchess to RMM-DIIS which is faster. And then VeryFast does all
electronic minimizations in RMM-DIIS. Has anyone used any of these
different algorithms successfully?

4.) EDIFFG and EDIFF: I am simply wondering what these values should
be during the final step in production runs. EDIFFG can be negative
meaning the calculation is stopped when the forces are less then
EDIFFG. EDIFFG should usually be 10 X EDIFF. Would a values of EDIFF =
10^-5 and EDIFFG=10^-4 be reasonable?

5.) ENCUT: Usually I have ENCUT set at 1.3Xmax(ENMAX in POTCAR),
during volume optimization. This ensures that I don't have pullay
stress errors. Would this be overkill for all relaxation runs
(especially the first few)?

Thank you for your help and replies and for any other info you can
shed onto USPEX and VASP production runs.

~Daryn