This thread is locked.Only browsing is available.
Top Page > Browsing
tight binding parameter extraction
Date: 2011/01/11 08:22
Name: J. Chang   <>


I am trying to use openmx code to extract tight binding parameters by setting up MLWF.

I have a question in the required time to get tight binding parameters.
Because my target material has five basis atoms and I assigned 4 orbitals on each atom, it takes a lot of time to do tight binding extraction calculation.
I did it in a serial mode, and took 7 or 8 days in my desktop.
So, I tried to run it in a parallel mode ( in this case 5 cpus are assigned, since the number of atoms are five).
But I could not see much time reduction to do the same calculation in the parallel mode.

My question is that
Does MLWF calculation run only serially even when I try to run it in parallel mode? or is it also parallelized?

Please give me some comments.

Page: [1]

Re: tight binding parameter extraction ( No.1 )
Date: 2011/01/12 02:39
Name: T.Ozaki


The computational time of TB parameters for such a small system is less than
a few minutes in general even if we do a serial calculation. The possible reason
for your case can be improper compilation or setting for parameters related
to the WF generation.

Have you tried to run the examples for the WF generation?

Best regards,

Re: tight binding parameter extraction ( No.2 )
Date: 2011/01/13 02:07
Name: J. Chang  <>

Thanks Dr. Ozaki,

I did calculations of wf examples which are provided by openmx itself.
Most of them does not take much time.
But in my case, it took more than a week.

My material is Bi2Se3.(5 basis atoms)
I assigned sp3 orbitals for each atoms, so total 20 orbitals.
Additionally, I considered spin up and spin down orbitals, so final number of orbitals are 20x2=40.

Could this number of orbitals lead to much longer running time?
If not, I think I made some mistakes.

Please give me advices.

J. Chang

Page: [1]