[Pw_forum] questions on vc-relax

[Axel Kohlmeyer]

On Fri, 1 Dec 2006, Guoying Gao wrote:


dear gao,

GG> Dear everyone,
GG>       I am doing structure optimization and set
GG> "calculation=vc-relax", I'm confused  that it doesn't converge no
GG> matter what I do. The following is what I have done:

what does not converge. the system size/shape, the atom positions,
or the wavefunction? please provide some sample input or at
least more details (atom types, atom count, plane wave cutoff).

GG>      nstep=100, I tried dt with the values of 30,50,75,90,100,150
GG>      Also, I do the above things with different pseudopotentials, the
GG> results are still the same.

please provide a more detailed description of th 'symptoms'. 

GG>      I also read the truobleshooting in version of 3.2 of  manual and
GG> Pw_forum, but I didn't find the suitable answers to my questions. The
GG> only prossible method is to increase variables "etot_conv_thr" and
GG> "forc_conv_thr", however, I don't know whether this will affect the
GG> accuracy of calculating the properties of the structure.

there have been multiple discussions on how to do this kind of
calculation well, so please check the mailing list archives.
first of all, you have to make sure that your stress tensor is
reasonably converged with respect to the basis set, then you have 
to either use the 'constant cutoff'-scheme or raise the cutoff 
even further, so that you do not suffer from so-called pulay stress.
rather tight convergence of the wavefunction is a requirement, too.

it is probably best, you start experimenting with examples that are
known to work, e.g. the ones presented at (with useful advice):
http://www.vlab.msi.umn.edu/events/lecture.shtml

GG> Does anyone know the reason for this or the method to solve it?

since we don't know exactly, what your problem is, it is _very_
difficult to give any but very generic advice. whether this applies
to your system or not, is very hard to tell.

cheers,
  axel.

GG> Thank you in advance!
GG> 
GG>                         gao
GG> _______________________________________________
GG> Pw_forum mailing list
GG> Pw_forum at pwscf.org
GG> http://www.democritos.it/mailman/listinfo/pw_forum
GG> 

-- 
=======================================================================
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.