Learn how easy it is to sync an existing GitHub or Google Code repo to a SourceForge project! See Demo
I want a comparison between the energetics of Mn2NiSn in the cubic inverse Heusler and Heusler phases. I had previously done these comparison using a pseudopotential code (Quantun Espresso). For the inverse Heusler phase, the code (ELK) works fine. But, I get problems in calculating energies and magnetism in the Heusler structure. The Heusler unit cell has four atoms at (000), (0.5,0.5,0.5) and (0.25,0.25,0.25), (0.75,0.75,0.75). I had varied the lattice constants from 10.70 a.u. to 11.60. Up to 11.20 a.u. the scf solutions reached to a nearly zero individual and total magnetic moments state, while in the later cases, the results were as expected (~5 mu_B). The energies varies in a zig-zag fashion as a function of lattice parameters making me unable to locate the equilibrium lattice constant. I've used the same input file throughout the calculations. Please have a look at my input file below and tell me where I've done wrong……..
! the large magnetic local field on Ni is halved at every iteration
! fairly large number of empty states required for magnetic cases
! lattice vectors
0.0 0.5 0.5
0.5 0.0 0.5
0.5 0.5 0.0
3 : nspecies
'Mn.in' : spfname
2 : natoms
0.00 0.00 0.00 0.0 0.0 2.000 : atposl, bfcmt
-0.50 0.50 0.50 0.0 0.0 2.000 : atposl, bfcmt
'Ni.in' : spfname
1 : natoms
0.25 0.25 0.25 0.0 0.0 0.300 : atposl, bfcmt
'Sn.in' : spfname
1 : natoms
-0.25 -0.25 -0.25 0.0 0.0 0.000 : atpos , bfcmt
12 12 12
0.5 0.5 0.5
your general setup looks OK, but I think you have got the structure wrong. Elk uses relative coordinates in terms of the lattice vectors for the atomic positions not Cartesian. Hence your -0.5 0.5 0.5 (relative) -> 0.5 0 0, while 0.5 0.5 0.5 -> 0.5 0.5 0.5 also in Cartesian coordinates. You want the latter, right?
The wrong structure could have lead to strange neighbour distances and problem with semi-core states which would explain your unstable energies.
I hope this helps, good luck!
Sorry my bad.
0.5 0 0 and 0.5 0.5 0.5 are of course equivalent for the fcc lattice.
Then I would suggest that you start with your non-magnetic solutions and put on the magnetic field again, i.e. essentially just put task=1 and see whether you do not converge to a magnetic solution.
The moments sometimes vanish when the external magnetic field is zero. With the value of reducebf that you are using, maybe the solution does not converge to a magnetic one. I generally start from a large external field and, once a magnetic solution is found, reduce the external field manually until the moments and total energy do not change much.