My Community

Hello VASP users,
I am trying to optimize a surface-dye complex (total 140 atoms) with PAW potentials using following input in INCAR file:

PREC = Accurate ! medium precision
EDIFF = 1e-4 ! electron relaxation criteria
EDIFFG = 1e-3 ! ion relaxation criteria
ENCUT = 400 ! Energy cutoff
ISIF = 2 ! internal coordinates relaxation
IBRION = 2 ! CG
POTIM = 0.1
ICHARG = 2 ! superposition of atomic charge densities as initial density
ISYM = 2 ! memory conserving symmetry calculation
NSW = 300 ! maximum ion relaxation steps

ISMEAR = 0 ! Gaussian smearing
SIGMA = 0.01 ! width of smearing in eV

ALGO = Normal ! mixed algorism with Davidson and RMM-DIIS

LWAVE = F ! Save wave function
LCHARG = F ! Save charge densities
LREAL = auto ! Real space projection
SYMPREC = 1e-6


The optimization was running for 4 days on 64 processors, But is is not showing convergence.

Do I need to make some corrections in the INCAR file?
Can we freeze surface atoms of the complex?

Any suggestion would be of great help.

Sincere regards,

Saurabh

How many ionic steps do you have?
Do you get any errors?(in standard output, not OUTCAR--> sigma could be to small then)
Do the number of electronic steps decrease per ionic step of do they remain the same(and how many? 60 or 10-20, If you get 60 then there is a problem with the electronic convergence already)

EDIFF might be a bit small (I tend to use 1.0E-5 and 1.0E-6, even for 200 atoms/cell)

You can try a different ALGO (use the standard one instead of the RMM-DIIS)

wrt freezin atoms...that depends on how strong your interactions are, how easily is your surface modified?

can you post some output?(OSZICAR) that might help identify some problem.

Danny

PS: us a viewer like p4vasp to check the change of the forces during the relaxation, and how the atoms move, this might also give you a clue as to the cause of your problem...it could be that your starting-guess is just really bad...or that you systems goes back and forth between two geometries...