[Wien] BFGS minimization

[Peter Blaha]

I've been testing this version.

Partially it worked very well, particularly for small test systems.
However, for bigger tests it got stuck and even moved atoms
into the wrong direction.

Thus, at the moment I would not like to release it in public. However,
interested (and experienced !!) users can ask me directly and I send it to
them for testing and further evaluation.

Experienced means: you must be able to monitor a minimization and judge
if it works or if it does not!!


> I'm planning to do some optimizations for structures with vacancies.  A good minimizer would be very useful for me.  I'd be happy to use and test your approach.
> Good that someone is finally trying to make wien optimization better!!
>
> Kevin.
>
>
> 	-----Oorspronkelijk bericht-----
> 	Van: L. D. Marks [mailto:L-marks at northwestern.edu]
> 	Verzonden: di 2/24/2004 2:59
> 	Aan: 'wien at zeus.theochem.tuwien.ac.at'
> 	CC:
> 	Onderwerp: [Wien] BFGS minimization
>
>
>
> 	I have coded a somewhat more robust BFGS miminizer from the
> 	PORT library (http://www-out.bell-labs.com/project/PORT/)
> 	into WIEN2K. It seems to work fine, and if anyone wants to
> 	test it (I have limited computer resources) contact me
> 	directly. It is currently implimented using the energy
> 	and forces (-fc), but I wonder if that is the right approach
> 	to use in all cases, particularly if approximations (for
> 	speed) are made with RKMAX/k-points etc. I'd be interested in
> 	comments (since I have limited experience) about how useful
> 	minimization on ||Force||, either full or just the HF forces
> 	might be, particularly if these converge faster with approximate
> 	RKMAX than the energy.


                                      P.Blaha
--------------------------------------------------------------------------
Peter BLAHA, Inst.f. Materials Chemistry, TU Vienna, A-1060 Vienna
Phone: +43-1-58801-15671             FAX: +43-1-58801-15698
Email: blaha at theochem.tuwien.ac.at    WWW: http://info.tuwien.ac.at/theochem/
--------------------------------------------------------------------------