Exchange Tensor - Vampire Unit Cell File

Michail Tzoumanekas

unread,

Dec 22, 2021, 5:51:45 AM12/22/21

to TB2J

Dear He Xu,

Thank you for your software! I am using TB2J for the calculation of the exchange parameters to run atomistic simulations with Vampire. I have contacted Non - Collinear DFT calculations with Spin-Orbit Coupling for the nanoparticle L10-CoPt with Quantum Espresso, using fully relativistic pseudopotentials for both atoms and the rotated structure method keeping the magnetizations along the z-direction, as you suggest.

- The command I am using for the rotation of the structure:

TB2J_rotate.py POSCAR --ftype POSCAR

- The command I am using for wann2J.py is, for the three components(x,y,z):

wann2J.py --spinor --groupby spin --posfile CoPt.scf --efermi 20.2119 --kmesh 8 8 8 --elements Co Pt --prefix_spinor CoPt --rcut 3.7266309999999998 *

* I am using rcut to have interactions in the range of one unit cell

- The command to merge the results:

TB2J_merge.py x y z(Folders) --type structure

I am attaching the results to have a look at them. My questions are as follows:

1. To the vampire material file: Why I am getting different atomic magnetic moments, than what I expected? (scf.out : Co_mz= 2.05 μΒ, Pt_mz=0.41 μΒ)

2. To the vampire unit cell file: While the Jani and DMI components have been computed in the exchange out file, why does the unit cell file contains only an isotropic exchange tensor?

3. How can I calculate the Tensor that describes the exchange interactions including the DMI? I am expecting the exchange tensor to have this form:

Jxx Dz,ij -Dy,ij

Jij = -Dz,ij Jyy Dx,ij

Dy,ij -Dx,ij Jzz

Thank you in advance,

Best regards

Michail Tzoumanekas

unread,

Dec 22, 2021, 5:52:41 AM12/22/21

to TB2J

rotated_structure.zip

Xu He

unread,

Dec 23, 2021, 3:25:23 PM12/23/21

to TB2J

Dear Tzoumanekas,

Thanks for the detailed command and files. I have some comments:

- The rcut is probably too small. For a metallic system the exchange can extend to several nanometers.

- The Wannier functions are not well localized as you can find from the Wannier output files . That is why the magnetic moment is wrong.

WF centre and spread   31 ( 1.863994, 1.315466, -1.205678 )    12.56134050
WF centre and spread   32 ( 1.866236, 1.309839, -1.171829 )    12.59647096
WF centre and spread   33 ( 1.528343, 1.222743, -1.181710 )    22.70633825
WF centre and spread   34 ( 1.538447, 1.327115, -1.034018 )    22.48222526
WF centre and spread   35 ( 1.855759, 1.145606, -0.663165 )    19.34697761
WF centre and spread   36 ( 1.854634, 1.171508, -0.676266 )    19.39655732
WF centre and spread   37 ( 1.821886, 0.661574, -1.160920 )    22.53958937
WF centre and spread   38 ( 1.785259, 0.690695, -1.325197 )    22.87838731
WF centre and spread   39 ( 1.353610, 1.049842, -1.375414 )    28.80178398
WF centre and spread   40 ( 1.350688, 1.221080, -0.987940 )    29.22119234
WF centre and spread   41 ( 1.298139, 1.164439, -1.315030 )    22.26557383
WF centre and spread   42 ( 1.334154, 1.085701, -1.325389 )    22.20455544
WF centre and spread   43 ( 1.832940, 1.272762, -1.320670 )    16.67039868
WF centre and spread   44 ( 1.834305, 1.287645, -1.329596 )    17.07480942

- the --groupby spin option probably need to be checked. In most DFT code the --groupby orbital is used.

- The vampire format outputted from TB2J is probably outdated. I am slow in updating the Vampire interface as I am not an active user of it. I will update it when I have time. I'd be happy if you have a list of incompatibilities between the two code.

- The tensor you write is correct. I am not sure if the Jxx, Jyy, Jzz are only from the anisotropic part, or does it include the isotropic one? The latter is right.