[Wien] problems with supercells

[Torsten Andersen]

Dear Mr. Torres,

Sergio Galindo Torres wrote:

> I'm studying the band structure of the Si crystal with Al impurities 
> qith Wien 98.7 package and I have tree questions about it:
>
> The last week i send a mail asking for a problem that I have in the 
> SCF cycle. It stoped after a few iterations. The solution that you 
> gave me was to remove the broyd and scf files, change the broyden 
> factor from 0.4 to 0.1 and run x dstart and run_lapw again. The Scf 
> cycle then stoped after just one iteration and the convergence wasn't 
> reached yet. Again no error messages were displayed in any of the 
> error and ouput files. So, what can i do? 

And there is nothing at the end of your case.dayfile? Or at the end of 
unit 6 (redirected stdout)?

>
>
> My other question is conserning the physical problem: In the structure 
> file, How do I set the RMT fpr the Si and the Al?. I heard that there 
> is a problem with the difference of the atomic radii when one works 
> the impurity problem in a LAPW based program. 

Set them about the same size, if possible (within 10% difference). The 
main problem here, as far as I know, is that you generate a lot of 
planewaves by having them too different, thus increasing your 
computation time.

>
>
> My last question is about performance of the calculation. I work with 
> a supercell of 64 atoms and it takes 3 days for just one iteration 
> (even when I work in a good platform: Intel Xeon 1.5GHz with 2GB RAM). 
> I compiled the program with the -O3 option for the f77 compiler and 
> force an iterative diagonalization with the -it flag in the run_lapw. 
> Should I reduce the size of the matrices in the Global parameters?. 
> You also advised in the manual to compile with and option depending of 
> the particular arquitecture. What is this option for the platform that 
> i just mentioned.

Well, I also have cases that takes several days for an iteration on good 
machines. Unless your parameters are extremely large, it should not make 
a big difference by adjusting them (and upgrading to Wien2k elimitates 
the dimensioning almost completely). However, you could try to decrease 
RKMAX in case.in1(c) and GMAX in case.in2(c) to lower the calculation time.

Best regards,
Torsten Andersen.

>
> Thanks
>
> _________________________________________________________________
> Charla con tus amigos en línea mediante MSN Messenger: 
> http://messenger.yupimsn.com/
>
> _______________________________________________
> Wien mailing list
> Wien at zeus.theochem.tuwien.ac.at
> http://zeus.theochem.tuwien.ac.at/mailman/listinfo/wien
>
>

-- 
Dr. Torsten Andersen
Department of Physics, Condensed Matter Theory Group, Uppsala University
UU-WWW: http://www.fysik4.fysik.uu.se/    TA-WWW: http://deep.at/myspace
New: Summer School 2003: http://www.fysik4.fysik.uu.se/~thor/school.html