Questions about MAGMOM-tag / CHGCAR file

Message

boris · #1 Post by **boris** » Wed Jan 16, 2008 2:30 pm

Hi.

I'm witnessing something strange with the CHGCAR file and the MAGMOM-tag in my calculations.

I'm trying to get a specific antiferromagnetic configuration for my system. To make the convergence easier, I'm first calculating the system 'from scratch' with no magnetisation (ISPIN=1, ISTART=0, ICHARG=2) then I'm using the resulting CHGCAR file to restart the calculation with magnetisation (ISPIN=2, ISTART=0, ICHARG=1).

The resulting CHGCAR file should contain information about the magnetisation density (since it was calculated with ISPIN=2 + MAGMOM-tag). Thus, if I want to use it for a continuation job, I have to comment the MAGMOM line in the INCAR file, as requested in the VASP User Guide.

When I'm doing this, the result is strange because the system converges towards a non magnetic solution. All magnetic moments on every atom is zero, in each component s,p,d or f :

magnetization (x)

# of ion s p d f tot
------------------------------------------------
1 0.000 0.000 0.000 0.000 0.000
...
96 0.000 0.000 0.000 0.000 0.000
------------------------------------------------
tot 0.000 0.000 0.000 0.000 0.000

If I want my system to be magnetic (restarting from this CHGCAR file), I have to specify MAGMOM in my INCAR file, but whatever the values I write, VASP does not seem to take it into account.

How come the CHGCAR file resulting from a "(ISPIN=1 ICHARG=2) then (ISPIN=2 ICHARG=1)" operation is non magnetic ?

My second question deals with the MAGMOM-tag : How does VASP use the values of magnetic moments ? Are they just an indication about the desired ground state magnetic configuration ? I'm using GGA+U and it seems that my system's final state strongly depends on the values in the MAGMOM-tag. Shouldn't VASP converge to an unique magnetic configuration whatever the values given in the MAGMOM-tag ?

Thanks in advance for your answers,

Kind regards.
<span class='smallblacktext'>[ Edited ]</span>

#2 Post by **admin** » Mon Jan 28, 2008 12:20 pm

please check (from the total energies) if the ground state of your system is magnetic or not (vasp converges to the electronic ground state)
-- to obtain high-spin solutions, MAGMOM has to be chosen quite large sometimes, please note that the read MAGMOM only serves as an initial guess for the magnetic moments)
-- for PW91 potentials, please set VOSKOWN=1
-- please set LMAXMIX appropriately

boris · #3 Post by **boris** » Mon Jan 28, 2008 12:48 pm

Thank you for your answer.

The ground state of my system is antiferromagnetic. As it is sometimes hard to get the desired magnetic solution, I'm using the procedure written in the VASP guide : MAGMOM-tag

I'm using a generated non magnetic CHGCAR. For the continuation job, I'm setting :

ISPIN = 2
ISTART = 0
ICHARG = 1
MAGMOM = -1 1 (for instance)

This leads me to the desired magnetic ground state.

But let's say I want to use the generated CHGCAR to restart the same calculation with improved convergence parameters. I need to set :

ISPIN = 2
ISTART = 0
ICHARG = 1

And I need to comment the MAGMOM-tag because I'm restarting from a (supposed) magnetic CHGCAR file.

When I'm doing this, magnetic moments on each atom are zero, and the total energy of my system is higher because it is not the correct magnetic ground state.

I don' t understand.

I'm using PBE potentials and my LMAXMIX = 6 since I'm dealing with f-elements (uranium).

shihongting · #4 Post by **shihongting** » Mon Feb 18, 2008 11:59 am

Hello, I have a question and want to ask you.
Are you calculating Uranium ?
I am investigating Nd which has the a-U phase under high pressures. How do you optimize the lattice contants a, b and c, as well as internal parameter y ?
I used the setting :
IBRION = 2
ISIF = 4
but the the structure relaxes to the ideal configuration, c/a = 1.63, b/a = root(3) , y = 1.67, which is far from the experimental data.

My email: hongting.shi@fysik.uu.se

Looking foward to help.

#5 Post by **admin** » Wed Feb 20, 2008 8:06 am

please make sure that your cutoffs and precision are sufficiently high. Also, please note that different XC-types usually yield (slightly) different results in V and c/a, therefore it may be worthwile to test which PP gives the best agreement with experiment.