ENCUT convergence check increasing rather than decreasing

[rajeshraju]

Hi,

I am woking on N-doped graphene (49 C and 1 N) and did convergence check to see the optimal (or the best) K and Encut values to be chosen for the minimization. I got the following numbers which show some ups and downs in the total energy written in the OUTCAR (energy sigma->0). Any help to figure it out? I presume, higher K-point and EnCut should provide better values.

400 -459.86236853 -459.8575553 -459.8566265
450 -459.52324679 -459.5181853 -459.5175918
500 -459.41923826 -459.4143076 -459.4138069
550 -459.43005966 -459.4251268 -459.4245913
600 -459.53038956 -459.5255936 -459.5249424

I read in the forum something we should substract kinetic energy erorr from the total energ before checking the values? I got the following the values for Kinetic Energy Error. That means I have to multiply 0.0057 x 49 and 0.0103 one time for each other. This was not clear in the forum answer.
kinetic energy error for atom= 0.0057 (will be added to EATOM!!)
kinetic energy error for atom= 0.0103 (will be added to EATOM!!)


the INCAR for Single point energy on the optimized geometry is shown below to check whether there is any error or not

# Basic setup:
SYSTEM = N-doped Graphene
ISPIN = 2
NCORE=24
KPAR=5
LSORBIT = .FALSE.
# not using Hubbard U

# Accuracy controls:
# PREC = High
ENCUT = 600

# Electronic loop controls:
ALGO = Fast # Fast (Davidson + RMM-DIIS)
EDIFF = 1E-7
NELMIN=15
NELM=60
#TIME = 0.400000
ISMEAR = 2 # 2nd order Methfessel-Paxton method
SIGMA = 0.2 # eV
# NBANDS = 144

# Relaxation control:
IBRION = -1
NSW = 0
POTIM = 0.2
ISIF = 2 # Ions Cell Volume
EDIFFG = 1E-5

# Properties:
LCHARG = .TRUE.
LWAVE = .TRUE.
# LVTOT = .FALSE. #Default False
# LVHAR = .FALSE.

I appreciate your help

Thanks
Rajesh

[alex]

Hi rajeshraju,

you are right with your assumptions. You have to remove the summed up 'kinetic energy error for atom' energies for all atoms. This extra energy was introduced to account for the not-complete basis set and extrapolate for the error due to that. Hence the name.
Be careful, these values are ENCUT - dependend. If everything is fine afterwards, your energies should decrease monotonic. If not, your calculation went wrong.

Cheers,

alex