Non-collinear magnetism and symmetries

Total energy, geometry optimization, DFT+U, spin....

Moderator: bguster

Locked
mleitner
Posts: 2
Joined: Thu Oct 06, 2011 4:30 pm

Non-collinear magnetism and symmetries

Post by mleitner » Thu Dec 14, 2017 4:19 pm

Hello,
I tried to compute about the simplest conceivable non-collinear antiferromagnetic system, i.e. fcc Mn. The parameter file is added below. However, it dies with the complaint "out of memory in cg", which I can understand, given that it tried to allocate 7 GB of memory. The problem lies in the fact that it determined the space group to be P1, that is, no symmetry whatsoever, and thus wants to compute 1728 k-points. I would say that the space group should be at least P-4 3 m (#215), needing only 84 k-points, and actually it should even be a primitive unit cell when the orientation of the spinor is transformed along with the positions.
So is there some switch I have overlooked so that abinit recognizes the correct symmetry and thus is able to treat this problem, or is this a bug in the symmetry finder, or is this a kludge because the handling of symmetric non-collinearities is not implemented?

Code: Select all

#Definition of occupation numbers
occopt 4
tsmear 0.005

#Definition of the unit cell
acell 3*6.765
rprim  1 0 0   # FCC primitive vectors (to be scaled by acell)
       0 1 0
       0 0 1

#Definition of the atom types
ntypat 1
znucl 25

#Definition of the planewave basis set
ecut  25         # Maximal kinetic energy cut-off, in Hartree
pawecutdg 50.

#Exchange-correlation functional
ixc 11             # GGA PBE

#Definition of the k-point grid
shiftk 0. 0. 0.
kptrlatt
    12 0 0
    0 12 0
    0 0 12

#Definition of the SCF procedure
iscf 17
nstep 50          # Maximal number of SCF cycles
tolvrs 1e-12

natom 4
typat 1 1 1 1

xred
 -0.25 -0.25 -0.25
 0.25 0.25 -0.25
 0.25 -0.25 0.25
 -0.25 0.25 0.25
 
spinat   0.5 0.5 0.5
         -0.5 -0.5 0.5
         -0.5 0.5 -0.5
         0.5 -0.5 -0.5
nsppol   1
nspden   4
nspinor  2

ebousquet
Posts: 469
Joined: Tue Apr 19, 2011 11:13 am
Location: University of Liege, Belgium

Re: Non-collinear magnetism and symmetries

Post by ebousquet » Wed Dec 20, 2017 10:56 am

Dear mleitner,
The symmetry routine is not yet ready for non-collinear magnetism... Sorry about that, we should add a warning in the code.
Although, if the GGA gave you problems of convergence, try also LDA. You might also have to reduce the diemix and diemixmag flags to improve convergence too.
We are working right now on the non-collinear par of Abinit to make it more efficient.
Check also if you really need a 12x12x12 kpoint grid for this supercell. I've tested the anisotropy of Fe-BCC and found that it was converged at 0.01 micro eV with a grid of 14x14x14 in the small FM unit cell, thought Mn is different...
Best wishes,
Eric

Locked