My Community

I have several questions:

1. about OPTIM tag, if i use IBRION=2, do I NEED to set OPTIM value (is there a default value)? Is there a rule to choose the value? I usually use 0.5, but i'm not sure if the inproper value causes slow convergency.

2. I remember read it somewhere that EDIFFG usually is 10 time of EDIFF. I wonder what happens, say, if EDIFF=1E-5, EDIFFG=-5E-2?

3. I ran a calculation with ISPIN=1 and then copied converged geometry as POSCAR, swithch ISPIN=2. I thought it would save lots of calculation time. However, for ISPIN=2 calculation, it took 2 hours to finish the first ionic step. Is this normal? here's INCAR setups:
ISMEAR = 0
ENCUT = 500
ENAUG = 660
NELMIN = 6
NELM = 200
EDIFF = 5E-3
EDIFFG = -5E-2
NSW = 3
IBRION = 2
ISPIN = 2

LORBIT = 10
LDAU=.TRUE.
LDAUTYPE = 2
LDAUL = -1 -1 3
LDAUU = 0.0 0.0 5.0
LDAUJ = 0.0 0.0 0.0

LMAXMIX = 6
ADDGRID = .TRUE.

AMIX = 0.2
BMIX = 0.0001
AMIX_MAG = 0.8
BMIX_MAG = 0.0001

HERE'S ONE MORE:
4. I understand that for a IBRION=2 calculation, NSTEP:
1 initial positions
2 trial step
3 corrector step, i.e. positions corresponding to anticipated minimum
4 trial step
5 corrector step----
so, say, if my linux account has time limitation so I can only run a calculation for 2 hours, but I don't want my calculation to be forced to terminate because it reaches time limit. Instead, I want to stop calculation before it reaches time limit so that there would be WAVECAR written out for next run by setting small NSW. Would it be more efficient to set NSW to be 3, 5, 7, 9 ----?

Thanks a lot!

1) EDIFFG is 10*EDIFF by default if not set explicitely, you are right (this implies that the energy differences are taken as ionic convergence criterium then)
2) it depends on the size of your system how long 1 ionic step takes, 2 hrs do not seem strange to me for large cells with high precision, ...
3) simply check how long one ionic step takes and set NSW to the number of steps which can be done in 2 hrs (rather round off to be sure).