My Community

[quote="alex"]Post the POSCAR, please.
And do a
grep TIT POTCAR

Thx

alex[/quote]TbDyFe
7.347
1 0 0
0 1 0
0 0 1
2 6 16
Direct
0.0000000000000000 0.0000000000000000 0.0000000000000000 #Tb2
0.5000000000000000 0.5000000000000000 0.0000000000000000
0.0000000000000000 0.5000000000000000 0.5000000000000000 #Dy6
0.5000000000000000 0.0000000000000000 0.5000000000000000
0.7500000000000000 0.2500000000000000 0.7500000000000000
0.2500000000000000 0.2500000000000000 0.2500000000000000
0.2500000000000000 0.7500000000000000 0.7500000000000000
0.7500000000000000 0.7500000000000000 0.2500000000000000
0.6250000000000000 0.6250000000000000 0.6250000000000000 #Fe16
0.6250000000000000 0.1250000000000000 0.1250000000000000
0.1250000000000000 0.6250000000000000 0.1250000000000000
0.1250000000000000 0.1250000000000000 0.6250000000000000
0.3750000000000000 0.8750000000000000 0.1250000000000000
0.3750000000000000 0.3750000000000000 0.6250000000000000
0.8750000000000000 0.1250000000000000 0.3750000000000000
0.8750000000000000 0.6250000000000000 0.8750000000000000
0.1250000000000000 0.3750000000000000 0.8750000000000000
0.6250000000000000 0.3750000000000000 0.3750000000000000
0.3750000000000000 0.6250000000000000 0.3750000000000000
0.8750000000000000 0.3750000000000000 0.1250000000000000
0.8750000000000000 0.8750000000000000 0.6250000000000000
0.3750000000000000 0.1250000000000000 0.8750000000000000
0.1250000000000000 0.8750000000000000 0.3750000000000000
0.6250000000000000 0.8750000000000000 0.8750000000000000[/blockquote]

the result of grep TIT POTCAR are :
TITEL = PAW_PBE Tb 23Dec2003
TITEL = PAW_PBE Dy 23Dec2003
TITEL = PAW_PBE Fe 06Sep2000

Looking forward to your reply:)

Hi, I tried with a smaller k-point set first and different BMIX/BMIX_MAG settings from 0.001 to 0.1. Always convergence within 35 cycles or so. For sanity I'll also do the 5x5x5 mesh, but this takes overnight.

It looks like your system behaves well and you do not need such harsh modification of the standard settings.

Cheers,

Alex

My INCAR:

SYSTEM = Laves stuff

Electronic minimisation
GGA = PE
PREC = Accu
EDIFFG = -0.005
ENCUT = 300.
ISPIN = 2
# AMIX = 0.2 ! linear mixing parameter
# BMIX = 0.1 ! cutoff wave vector for Kerker mixing scheme
# AMIX_MAG = 0.8
# BMIX_MAG = 0.1
NELMIN = 4
NELM = 500 ! Maximum number of electronic SC steps
NELMDL = -20 ! Number of NON-selfconsistency steps (negative value: delay only for the start-configuration)
LMAXMIX = 6 ! PAW flag
MAXMIX = 50

PBE+U calculation: ! types & order: O Si Eu Al P
LDAU = .TRUE. ! Switches on the L(S)DA+U
LDAUTYPE = 2 ! Dudarev's approach to LSDA+U (Default)
LDAUL = 3 3 -1
LDAUU = 6.00 6.00 0.00
LDAUJ = 0.00 0.00 0.00
LDAUPRINT = 0 ! Controls verbosity of the L(S)DA+U module, 0: silent

DOS related values
SIGMA = 0.1
ISMEAR = 0 ! Gaussian smearing for big cells

Do not write huge files
LCHARG = .FALSE.
LWAVE = .FALSE.

Performance options
NPAR = 2 ! adjust to number of processors
ALGO = N

[quote author= Laves stuff

Electronic minimisation
GGA = PE
PREC = Accu
EDIFFG = -0.005
ENCUT = 300.
ISPIN = 2
# AMIX = 0.2 ! linear mixing parameter
# BMIX = 0.1 ! cutoff wave vector for Kerker mixing scheme
# AMIX_MAG = 0.8
# BMIX_MAG = 0.1
NELMIN = 4
NELM = 500 ! Maximum number of electronic SC steps
NELMDL = -20 ! Number of NON-selfconsistency steps (negative value: delay only for the start-configuration)
LMAXMIX = 6 ! PAW flag
MAXMIX = 50

PBE+U calculation: ! types & order: O Si Eu Al P
LDAU = .TRUE. ! Switches on the L(S)DA+U
LDAUTYPE = 2 ! Dudarev's approach to LSDA+U (Default)
LDAUL = 3 3 -1
LDAUU = 6.00 6.00 0.00
LDAUJ = 0.00 0.00 0.00
LDAUPRINT = 0 ! Controls verbosity of the L(S)DA+U module, 0: silent

DOS related values
SIGMA = 0.1
ISMEAR = 0 ! Gaussian smearing for big cells

Do not write huge files
LCHARG = .FALSE.
LWAVE = .FALSE.

Performance options
NPAR = 2 ! adjust to number of processors
ALGO = N

[/quote]


Hi, alex, sound amazing about your calculation result with small k mesh. What about k mesh 5*5*5. (mine is 7*7*7)

Let me make sure about one thing: are you using the pseudopotential (PAW_PBE) of Tb, Dy in which consider f electrons as valence electrons ?? (it means Tb and Dy have 19 and 20 valence electrons respectivily)
Thanks:)

Hi again, so 5x5x5 was fine, too. But you are right, I've used the ??_3 versions, but these are the only one I got. :-(
Are there more in the repository lately?

Cheers,

alex

[quote="alex"]Hi again, so 5x5x5 was fine, too. But you are right, I've used the ??_3 versions, but these are the only one I got. :-(
Are there more in the repository lately?

Cheers,

alex[/quote]

No wonder you get it converged, I already used the Tb_3 and Dy_3, too, and I also got it converged. But for Tb_3 and Dy_3 potential doesn't treat f electrons as valence, so we can do nothing about f electrons. I am very interested in the properties of f electrons, such as spin orbit coupling, lda+u to deal with strong correlated etc. So I have to use Tb and Dy potential which take f electrons as valence electrons.
You can ask vasp developer for the these potentials.

So, alex, you have any idea about how to get it converged by using these potentials(using lda+u, so coupling) ?
Thanks a lot:)