Use of JtoDel

This applies to mode = energies. The strategy used by JaSS is to remove configurations, but do this in a way to calculate as many exchange constants as possible. At some point one needs to decrease number of exchanges as well. This is done by removing exchanges corresponding to the longest exchange paths first. But the problem can be not due to these far exchanges and even not due to “bad” configurations, but because of “bad” nearest neighbor exchanges. These are exchanges, which are met the same number of times (and with the same signs) in all configurations and thus can’t be calculated by the total energy method. As a result JaSS decreases number configurations and exchanges until these “bad” exchanges disappear (from the expressions for the total energies) and one can easily left out with only a few Js even if number of magnetic ions is rather large.

The solution of the problem can be to study jass.out and in particular lines after

Search for possible magnetic configurations to calculate

JaSS prints out after these lines exchanges which can’t be calculated at the same time at each iteration of the configuration minimization. If there is a certain exchange, e.g., J2 is met all along minimization process, then this is the “bad” exchange. The suggestion would be to include this exchange to JtoDel and rerun JaSS. It well may be that now you will be able to calculate much more exchange constants.

 

Symmetry in GGA+U

There always can be interplay between spin and orbital degrees of freedom, e.g. via Kugel-Khomskii like coupling. Such a situation may realize in GGA+U calculations, where U part is orbital dependent. This is why a symmetry of electronic subsystem can be lower than crystalline. On the one hand, the use of the symmetry typically substantially decreases time of a calculation and improves convergence process, but on the other hand it may substantially restrict types of possible DFT+U solutions. This may lead in a situation when solution without any symmetry restrictions would have lower energy than the symmetric one. As a result exchange parameters can be wrong, if even in only one of the magnetic configurations orbitals degrees of freedom turned out be strongly coupled with spin ones. Thus, it is strongly recommended to perform DFT calculation without any symmetry, i.e., e.g., in VASP we suggest to use ISYM=0 tag.

 

VASP: from GGA to GGA+U

As it has been mentioned before, GGA+U is known to have numerous local minima (in contrast to GGA). Thus, one of good strategies could be to perform GGA alculation first and then switch on U . This can be easily done in JaSS. One needs to copy whole directory with all JaSS calculations (including all subfolders like 0, 1, 2 etc.). Modify INCAR file accordingly (add all U related tags and ICHARG=1), remove OUTCAR files from all subfolders (i.e. run ’rm */OUTCAR’ ) and execute jass. Then JaSS will start VASP with self-consistent GGA charge density, which strongly helps in convergence.

 

VASP: changing KPOINTS, U etc.

JaSS watches the parent directory and if you change number of k-points will overwrite KPOINTS files in all subdirectories and resubmit the calculations (since now they are considered as non-selfconsistent). However, JaSS can NOT trace changes in INCAR, since this file is modified by the program. Thus, in order to run JaSS with modified INCAR one needs to ’rm */OUTCAR’ and then run jass.

 

Wien2k: from GGA to GGA+U

Default type of JaSS calculation with using of Wien2k is GGA. If it is necessary to perform GGA+U calculation one has to put case.inorb and case.indm in the root directory together with case.struct and jass.inp. This action leads to appearing of -orb and -dm keys in the Asbatch - the file containing execution string for each magnetic configuration.

Back to top