"total drift" - tooooo big _Help needed

[prithwish]

Hi, I am doing a simulation for Y2Ti2O7 using PAW PBE ( Y_sv, Ti_sv, O) pseudopotentials. I have done the structural optimization in steps, i.e. initially only ionic relaxation, then volume relaxation and at last shape,volume & ionic relaxations.

For fist step, i.e. for IONIC relaxation , the structure is optimized. Both the conditions for EDIFF and EDIFF have been reached. In OUTCAR file, the forces on each atom is well below the specified cutoff, but "total drift" is surprisingly higher...

"total drift: -6.250822 -4.126612 4.307766"
Its toooo large.

Here, I am posting the INCAR file that I am using :
#########################################################
System = Y2Ti2O7

ISMEAR = 0 ! 1: Methfessel-Paxton order N; -1:Fermi-smearing.
SIGMA = 0.1 ! Smearing factor (in ev)
PREC = High ! sets cutoff and FFT grid
ENCUT = 600 ! planewave energy cutoff (in ev)[default taken from
! POTCAR file]
!ENAUG = 550 ! grep EAUG POTCAR

EDIFF = 1E-06 ! energy stoping criterion for electronic iteration
EDIFFG = -1E-3 ! energy stoping criterion for ionic Relaxation

ICHARG = 2 ! start from scratch(this flag determines how to construct the
! initial charge density. 0:calculate charge density from
! initial wave functions; 1:read the charge density from file
! CHGCAR and extrapolate from the old positions to new positions.
! in PAW and LDA+U method be careful (see manual); 2:take
! super-positions of atomic charges; 4: read potential file from
! POT file; +10: non self-consisitent calculation,it means that
!charge density will be kept constant during whole electronic
!minimization.
INIWAV = 1 ! 0:take "jellium wave functions"; 1: fill wave functions arrays
! with random nos (safest full-proof switch).
ISTART = 0 ! start from scratch.Initialize the wave functions according to
! the flag INIWAV
NELM = 60 ! maximum of 60 electronic steps.
!NELMIN = 5 ! minimum of 5 steps in each electronic SC
NELMDL = -12 !no update of charge for 12 steps.

LREAL = .TRUE. !( making the calculation fast)
!ALGO = Fast ! (making the calculation fast)
IALGO = 48
NSIM = 4 ! (making the calculation fast);blocked algorithm update;4 bands
! at a time
ISPIN = 2 ! 1:non-spin polarized calculations; 2:spin-polarized calculation
!MAGMOM = 16*1.5 16*1.5 56*1.5 ! (default: NIONS*1;for spin-pol cal a save default:expt mag mom
! *1.2 or 1.5)


LORBIT = 12

LWAVE = .TRUE. ! Print WAVCAR if .TRUE.
LCHARG = .TRUE. ! Print CHGCAR if .TRUE.
LVTOT = .TRUE.

ADDGRID = .TRUE. ! Set .TRUE. for PAW
LMAXMIX = 6
AMIX = 0.2
BMIX = 0.0001
AMIX_MAG = 0.8
BMIX_MAG = 0.0001

Ionic relaxation
NSW = 200 ! # of steps in optimization (default 0!)
ISIF = 2 ! 0: relax ions, 1,2:relax ions,calc stresses,
IBRION = 2 ! 1: quasi-NR, 2:CG algorithm for ions
NFREE = 10 ! number of DIIS vectors to save
POTIM = 0.5 ! reduce trial step in optimization

#Parallel Options
LPLANE = .TRUE.
LSCALU = .FALSE.
NPAR = 10 ! how many bands in parallel
NSIM = 4
~
SYMPREC = 1.e-3
##########################################################

I am doing the calculation using 4 x 4 x 4 K-points (Monkhorst - pack Scheme)

Can anyone please locate which of the parameters I need to tune in order to get very small drift ?


&|

[admin]

this really looks strange for a bulk system, because your input parameters look reasonable. Have you checked the remaining pressure on the cell?
You could try to use PREC=Accurate and increase SYMPREC (or leave it to the default value of 10-5)

[prithwish]

Hi, Thanks for the reply. I have checked for external pressures . Here I am listing the corresponding values for " total drift " and " external pressure " with ISIF = 2 as well as with ISIF = 3.

ISIF = 2

a) 1st ionic step :
total drift: -6.399568 -4.283848 4.280882
external pressure = 11.12 kB Pullay stress = 0.00 kB

b) 2nd ionic step:
total drift: -6.304052 -4.185524 4.293387
external pressure = 6.19 kB Pullay stress = 0.00 kB

c)3rd ionic step :
total drift: -6.260273 -4.140616 4.298631
external pressure = 4.19 kB Pullay stress = 0.00 kB

d) 4th ionic step:
total drift: -6.263487 -4.144015 4.298027
external pressure = 4.26 kB Pullay stress = 0.00 kB

2) ISIF = 3

a) 1st ionic step :
total drift: -6.263670 -4.144221 4.297979
external pressure = 4.30 kB Pullay stress = 0.00 kB

b) 2nd ionic step :
total drift: -6.273710 -4.150496 4.305215
external pressure = -6.91 kB Pullay stress = 0.00 kB

c) 3rd ionic step :
total drift: -6.269321 -4.148457 4.300578
external pressure = 0.02 kB Pullay stress = 0.00 kB

d) 4th ionic step :
total drift: -6.269810 -4.148923 4.300590
external pressure = 0.08 kB Pullay stress = 0.00 kB

Now, as you suggested, I have submitted two runs , the first one with PREC = Accurate and the second one with SYMPREC = default. Still its going on. I shall report it as soon as the runs are over.

[prithwish]

Hi, the problem of high value of " total drift " is solved.

Firstly, PREC = Accurate could NOT solve the problem

Then, I removed SYMPREC = 1.e-3 from command lines. I also replaced the cartesian coordinates in the POSCAR file with direct ( fractional) coordinates. It is because of the low precision of the cartesian coordinates , I introduced SYMPREC tag in the INCAR file, otherwise VASP was not able to detect the space group symmetry. But, that turned out toooo costly. Anyway, by using default value of SYMPREC and changing the POSCAR ( i.e. from cartesian to direct coordinates) both problems was solved. Now VASP can recognize the correct symmetry of the space group and total drift value was 0.0.

Thanks to " HEAD ADMIN " . Thanks to everybody.