magnetic moment for LiMn1.5Ni0.5O4

[mhankel]

Dear all,

I trying to get my head around calculations on LiMn1.5Ni0.5O4. I am looking at the barriers for Li hopping with and without oxygen vacancies. I have two different configurations, P432 and Fd3m. I use DFT-U for Mn and Ni.

For P432 I remove one Li in two different locations and then use CI-NEB to get the transition state for the Li hop. For Fd3m I first remove one oxygen or four oxygens and then once these are converged do I remove a single Li in two positions and then do a CI-NEB again.

However, even getting the right lattice constants proved rather difficult. I found that the total magnetic moment changed a lot for calculations where I varied a=b=c by hand (the lattice is cubic cell). So I tried different MAGMOM setttings in the INCAR file. I first used LORBIT=11 to write out the magnetic moments of all the atoms and then I used these, as well as magnetic moment from unpaired electrons in orbitals as well as magnetic moments found via a web search to see which combination (about 12 different ones) gives the lowest in energy. I found, for one fixed lattice constant, that most of the MAGMOM input lead to the same total magnetic moment (mag = 30). So I picked one and then ran the lattice optimisation again. But, yet again, I got a different mag for some of the lattice setting and annoyingly enough I got a different one (mag = 32) to the one before. So even with setting a MAGMOM in the INCAR file the total magnetic moment was not the ground state.

I then did a bit more playing around and found that mag = 28 seems the lowest and I used NUPDOWN = 28 to fix this. So I now at least have an optimised lattice.

However, I now have to remove 1 Li, or 1 O, or 4 O from the configuration and so have to figure out the new magnetic moment for all of these. Is there any way to set MAGMOM in the INCAR file to make the calculations more stable with regards to the magnetic moment and make sure that I get the ground state? Here is a typical INCAR file with one of the MAGMOM settings. I also use the Li_sv potential.


LPLANE = .TRUE.
ISTART = 0
ICHARG = 2
ENCUT = 600
PREC = Normal
EDIFF =1E-6
ALGO = Fast
ISMEAR = 0
SIGMA = 0.05
ISPIN = 2
LORBIT = 11
NSW = 1000
IBRION = 2
EDIFFG = -0.01

LDAU = .TRUE.
LDAUTYPE = 2
# Li Ni Mn O
LDAUL = -1 2 2 -1
LDAUU = 0 5.25 5.0 0
LDAUJ = 0 0 0 0

MAGMOM = 8*0.6 4*-1.5 12*3.5 32*-0.15

For fixing the total magnetic moment via NUPDOWN I used this (from OUTCAR of other runs Li and O have near zero magnetic moment while I just rounded the ones for Ni and Mn to make adding this up easier)

MAGMOM = 8*0.0 4*-2 12*3 32*0.0
NUPDOWN = 28

Thank you in advance for your help.

Marlies

[cpp6f]

No, all that MAGMOM does is set the magnetic moment used to determine the initial charge density. From my experience, running these kinds of calculations with GGA+U on systems with 3d TMs can be extremely challenging and you can converge to several magnetic ground states depending on the initial MAGMOM values. Once you start making vacancies, you could end up with localized charges on the metal atoms, for example the cations near an O/Li vacancy will have a lower/higher oxidation state and a different magnetic moment than the others. Is the original system metallic or insulating?