error in merging result

mailhexu / TB2J

a python package for computing magnetic interaction parameters

BSD 2-Clause "Simplified" License

67 stars 29 forks source link

error in merging result #60

Open lyonworton opened 4 weeks ago

lyonworton commented 4 weeks ago

When I calculated DMI and Jani, after merging, I found that the symmetry of the result obtained from the calculation did not correspond to the lattice symmetry, in addition, the change of the density of kmesh during the calculation process had a great influence on the merged result, and it never reached convergence, and the above made me doubt the reasonableness of the result.

mailhexu commented 3 weeks ago

Hello, Could you share the files so I can have a look? Best regards, HeXu

lyonworton commented 3 weeks ago

m.zip I rotated the structure and did the scf calculations at xyz in each direction to get the output file and performed the TB2J magnetic interaction calculations on the output file. kmesh was chosen to be 15151, 16161, 17171, etc. and the final merge results were found to be different and could not be converged, and the DMI symmetry and lattice symmetry were different, for example the same The DMI symmetry and lattice symmetry are not the same, for example, the same near-neighbor distance has different DMI values.

lyonworton commented 3 weeks ago

Also I'd like to ask why the value of J or D doesn't end up decaying to 0 with distance, but instead floats above and below 0, even though the decay is oscillatory for transition metals.

lyonworton commented 2 weeks ago

What is the problem, what is the relationship between DMI calculation and kmesh, I hope to get your reply!

mailhexu commented 2 weeks ago

Hello,

convergence w.r.t. k-mesh: For metals, it sometimes requires a very dense k-mesh to get the convergence. Note that the unit is meV and sometime the J/DMI/etc can be sub meV, to get to this level of convergence is indeed not easy. If you find the relative difference to be ~ 5%, I would say that is already well converged.
Lattice symmetry: there is indeed a slight difference in the J values for atom pairs with symmetry. This is likely due to numerical noise, (either in the DFT or the TB2J calculations). For the DMI, note that it is a (pseudo)vector and the rotation operation should be considered when comparing the values. We have developed a method to symmetrize the J and DMI by using the crystal symmetry, which will be soon released.
The decay of the J and D can be extended to long distance for this type of metals (which can be understood as the RKKY mechanism). Best regards, HeXu