[Pw_forum] Pseudoptential for Nd

Axel Kohlmeyer akohlmey at cmm.chem.upenn.edu
Fri May 9 21:39:52 CEST 2008


On Fri, 9 May 2008, Javier Antonio Montoya wrote:


javier,

JAM> Thanks a lot, I would really appreciate if you can find your Pr pseudo, I am
JAM> willing to take that risk. Same regarding the other ones, in case somebody
JAM> already have them.  The matter is that experimentally we already have the
JAM> Raman and Infrared peak positions, plus the lattice constants, therefore, it
JAM> will be pretty easy to decide whether or not the results make some sense and

what paolo was pointing out is that you have a _fundamental_ 
problem. so even if you get a match, it will be by accident
and not by design. what would that prove? only that you got
lucky! yes, you want to match the experiment, but you want
to match it for right reasons.

i've seen this with uranium for example. it was giving reasonable 
results for some uranium oxide compounds at some lattice constant 
values but then it would just give complete crap at some others.

JAM> then, to use them to support some microscopic analysis of the physics that
JAM> is going on in our samples. Each material contains a single rare-earth
JAM> element but it also contains oxygen and iron, I think there isn't a big risk

so if you get crap for one element out of many (and i would not
be so certain about iron either, that can be pretty nasty, too),
how does that make your results better than if all elements
would be doubtful. the only difference is that the error will
be less obvious.

JAM> on using the ones available from the repository in PWscf.org for those other
JAM> two, therefore I just need to satisfy the NLCC, Van-US, PBE requirement also
JAM> for the rare-earths in order to make them match.

why? all you have to match is the functional! ...and as paolo had 
pointed out in some other mail, it looks like the vanderbilt type 
ultrasofts don't perform so well for "tricky" elements. it is 
definitely tricky to even get those potentials created with the 
vanderbilt code in a way they look reasonable to begin with.

cheers,
   axel.

JAM> 
JAM> Cheers,
JAM> 
JAM> Javier.
JAM> 

-- 
=======================================================================
Axel Kohlmeyer   akohlmey at cmm.chem.upenn.edu   http://www.cmm.upenn.edu
   Center for Molecular Modeling   --   University of Pennsylvania
Department of Chemistry, 231 S.34th Street, Philadelphia, PA 19104-6323
tel: 1-215-898-1582,  fax: 1-215-573-6233,  office-tel: 1-215-898-5425
=======================================================================
If you make something idiot-proof, the universe creates a better idiot.


More information about the Pw_forum mailing list