My Community

Hi,
I have some problems.
1) Bulk calculation. The system of calculation is LiCoO2, the XRD results of this crystal’s parameters are a=0.28166nm and c=1.408nm. However, the calculation results are a= 0.312nm and c=1.5967nm.The error is more than 10%, it is too big. But I do not know the reason. Please help me.
here is the INCAR file:
SYSTEM = LiCoO2:R-3m

ISTART = 0
ICHARG = 2
LPETIM=F
BMIX = 1.6 mixing parameter
AMIX = 0.06
EDIFF =1.E-4
ENCUT = 900 eV
Ionic Relaxation
NSW = 0 number of steps for IOM
IBRION = 2
EDIFFG = -0.02
POTIM = 0.1

DOS related values:
ISMEAR = 0 ; SIGMA = 0.2
Electronic Relaxations
IALGO =48 algorithm
LDIAG = T sub-space diagonalisation
LREAL = O real-space projection (0 or Auto)
ROPT = 136*1E-4
SGI Origin optimized values:
LPLANE = .TRUE.
NPAR = 4
LSCALU = .FALSE.

If you use NSW= 0, I think there is no need to specify IBRION , EDIFG and POTIM ; this might be not the problem. Second, reduce the mixing parameters . Use PAW-PBE ( test the Psedupotential ) , test the convergence of K-points . Then get E for different V; V (min) to V(max), and finally plot E vs V and fit it one of the equation of states from which you can obtain the minimum values of the lattice constats. I think this could help you



<span class='smallblacktext'>[ Edited Sun Apr 23 2006, 07:42AM ]</span>

des' response has made me realise that with that INCAR you're not doing any relaxation of the cell. Or indeed at all (NSW=0). Are you scanning over the lattice constants or something?

In addition to setting ISIF=3 and NSW NOT equal to 0, it is extremely important to set PREC=High, ENCUT to at least 1.5* max(ENMAX) of POTCAR, and to converge the k-points first in order to get accurate results for the unit cell geometry. Please have a look at accuracy.pdf at http://cms.mpi.uniive.ac.at/vasp-workshop for further information and a few figures showing the importance of good convergence of all these parameters.

On the PREC-tag, Kresse says in that presentation that Medium and High are no longer recommended. Instead he says to use Normal or Accurate (on p. 24). Could you please clarify? For example, why we should use High instead of Accurate?

Put some thoughts on DFT+U or PBE0/HSE0X ...

the cutoff parameters for the projection operators are chosen better if Accurate is used