Very slow self-consistency on the first ionic relaxation step

[lanli]

Hi all!

i relax a configuration of FeC60, where a Fe atom bonds to an isolated C60 molecule, using PAW potentials. Electronic SC-steps on the first ionic relaxation step are > 100. How can I speed up the self-consistency? if i increase SIGMA or ENCUT, will it help?

Here is my INCAR file:
System = Fe-C60
ENCUT = 300.0
NELM = 300
GGA = 91
ISMEAR = 1
SIGMA = 0.2
PREC = Accurate
LPLANE = .T.
LREAL = .T.
NSW = 100
IBRION = 2
ISPIN = 2
ISYM = 0
AMIX = 0.2
BMIX = 0.0001

[admin]

your input looks reasonable. Please check whether you have used the Gamma point only in KPOINTS (as appropriate for free molecules), it may improve convergence to use Gaussian BZ-integration (ISMEAR=0) and the Vosko-Wilk-Nusair ansatz for the parametrization of the free electron gas contribution of the LDA-XC part (VOSKOWN=1).
Sometimes, difficulties in the electronic convergence arise if the ionic configuration of the starting geometry is bad
[ Edited Mon Jun 09 2008, 03:30PM ]

[cwandtj]

Sometimes, it's very helpful if you keep your WAVECAR and start your next calculation with the WAVECAR as writing WAVECAR is one of the reason for slow convergence in first ionic step, especially for large system like yours.

CW