Convergence problem with VASP.5.2.11 for already converged (using VASP.4.6) geometry ???
Posted: Thu Aug 09, 2012 7:04 am
Dear Admin & VASP user,
It seems very strange that relaxed geometry which I obtained using VASP.4.6 (within 12 ionic and 12-25 electronic steps in each ionic step) does not converge within 120 steps if I use VASP.5.2.11 (parallel).
Due to memory issues I can not do DOS calculation on serial clusters. For parallel calculation I take same INCAR, KPOINTS, POTCAR . I use CONTCAR of serial version as POSCAR of parallel version. I wanted to obtain only CHGCAR so that I can run DOS calculation. But convergence is very poor. I am taking already converged geometry. But with VASP.5.2.11 (||) after 120 steps instead of 10^-5 it reaches upto 10^-1. Convergence problem usually appears when the initial geometry is not correct. Which is not the case here. Then ?
I'm not understanding what is causing problem. Please suggest me how to handle this problem. I'll be be very thankful to you.
======================================================================
INCAR
Start parameter for this run:
ISTART = 0
ICHARG = 2
Electronic Relaxation 1
PREC = ACCURATE
ENCUT = 500
IALGO = 38 algorithm
NELM = 120 number of electronic steps between two ionic steps
NELMIN = 0
NELMDL = 3
EDIFF = 1E-05
GGA = PE
ISPIN = 2
MAGMOM= 1 1 1 1 3.5 1 1 1 1 1 1 1 1 1 1 1
LORBIT = 11
ADDGRID = .TRUE
LMAXMIX = 4
INIMIX = 0
AMIX = 0.2
BMIX = 0.0001
AMIX_MAG= 0.8
BMIX_MAG= 0.0001
MAXMIX = -12
Ionic Relaxation
NSW = 60 number of steps for IOM #
NBLOCK = 1
KBLOCK = 5 inner block; outer block
IBRION = 1
ISIF = 2
ISYM = 2 ! 1 for us-pp, 2 for PAW data sets are used
IWAVPR = 1
LCORR = T Harris-correction to forces
EDIFFG = -5.E-03 stopping-criterion for IOM
POTIM = .75 time-step for ion-motion
DOS related values:
SIGMA = 0.1; ISMEAR = -5 broad. in eV, -4-tet -1-fermi 0 gaus
Electronic Relaxation 2
LDIAG = T sub-space diagonalisation
NBANDS = 128
=====================================================================
best regards,
PS
It seems very strange that relaxed geometry which I obtained using VASP.4.6 (within 12 ionic and 12-25 electronic steps in each ionic step) does not converge within 120 steps if I use VASP.5.2.11 (parallel).
Due to memory issues I can not do DOS calculation on serial clusters. For parallel calculation I take same INCAR, KPOINTS, POTCAR . I use CONTCAR of serial version as POSCAR of parallel version. I wanted to obtain only CHGCAR so that I can run DOS calculation. But convergence is very poor. I am taking already converged geometry. But with VASP.5.2.11 (||) after 120 steps instead of 10^-5 it reaches upto 10^-1. Convergence problem usually appears when the initial geometry is not correct. Which is not the case here. Then ?
I'm not understanding what is causing problem. Please suggest me how to handle this problem. I'll be be very thankful to you.
======================================================================
INCAR
Start parameter for this run:
ISTART = 0
ICHARG = 2
Electronic Relaxation 1
PREC = ACCURATE
ENCUT = 500
IALGO = 38 algorithm
NELM = 120 number of electronic steps between two ionic steps
NELMIN = 0
NELMDL = 3
EDIFF = 1E-05
GGA = PE
ISPIN = 2
MAGMOM= 1 1 1 1 3.5 1 1 1 1 1 1 1 1 1 1 1
LORBIT = 11
ADDGRID = .TRUE
LMAXMIX = 4
INIMIX = 0
AMIX = 0.2
BMIX = 0.0001
AMIX_MAG= 0.8
BMIX_MAG= 0.0001
MAXMIX = -12
Ionic Relaxation
NSW = 60 number of steps for IOM #
NBLOCK = 1
KBLOCK = 5 inner block; outer block
IBRION = 1
ISIF = 2
ISYM = 2 ! 1 for us-pp, 2 for PAW data sets are used
IWAVPR = 1
LCORR = T Harris-correction to forces
EDIFFG = -5.E-03 stopping-criterion for IOM
POTIM = .75 time-step for ion-motion
DOS related values:
SIGMA = 0.1; ISMEAR = -5 broad. in eV, -4-tet -1-fermi 0 gaus
Electronic Relaxation 2
LDIAG = T sub-space diagonalisation
NBANDS = 128
=====================================================================
best regards,
PS