Re: NEGF Convergence problem in Magnetic junction ( No.1 ) |
- Date: 2014/06/06 21:10
- Name: Artem <artem.pulkin@epfl.ch>
- First of all, you have a problem with first 20 "periodic" iterations. Since they are very cheap I would suggest you to increase NEGF.scf.Iter.Band and to try to reach convergence there. Also it does not look like you relaxed your structures.
At this point, it is not a problem of NEGF calculations.
|
Re: NEGF Convergence problem in Magnetic junction ( No.2 ) |
- Date: 2014/06/09 21:57
- Name: Umar <umarfarooqphysics@gmail.com>
- Dear Artem,
After trying on your tips. I am writing here .. I tried NEGF.scf.Iter.Band up to 300 it did not converge there also I also tried with higher range of mixing wait up to .6 but it did not work. You are absolutely right about the I did not relaxed my structure on OpenMx but i am using structure relaxed on FLAP (full potential method) which is noticeably different.
even if it is not relaxed energy convergence should be achieved ? is it not possible on openmx ? is it helpful to use more no of orbital bases set ? or using making optimized bases set ? or any other way to converge it on this geometry ? (because it is better for the constancy of the work )
any tip which can be helpful ?
thanks
Umar Farooq
|
Re: NEGF Convergence problem in Magnetic junction ( No.3 ) |
- Date: 2014/06/10 18:36
- Name: Artem <artem.pulkin@epfl.ch>
- Your basis set is quite rich, I would say. Depending on the properties you are looking for, you may want to decrease it. In my research, I took the basis set which represents well the band structure of my material. It is usually enough if you study transport properties.
Try to switch to default Pulay parameters and decrease the basis size. I am not a specialist here but try for example
C C5.0-s2p2 C_PBE13 Co Co6.0S-s2p2d2 Co_PBE13S Ni Ni6.0S-s2p2d2 Ni_PBE13S
And see if it works.
|