[CPMD-list] Geometry optimization and misrepresentation of the x-ray data...

[Axel Kohlmeyer]

On Fri, 25 Apr 2008, Pedro Nuno Simões wrote:

dear pedro,

how do you expect people to investigate you problem,
if you do not give them a chance to reproduce it???

it has become a _very_ bad habit that people post 
incomplete inputs. please keep in mind, that the
chances to get help are the higher the easier you
make it to answer. i for myself am getting increasingly
tired of having to play 20 questions with people to
get out the full story, and i noticed similar trends
with other people on other mailing lists. it is also
important to note that the ratio of people asking to
people answering is changing toward the asking side
and not only because of more people asking...

if you think it would be too much of a bother, 
just put your file(s) up for download somewhere.

based on the little information you are providing,
it looks like you have one or a combination of the 
following problems (in decreasing order of 
likelyhood, but remember this is pure speculation!):

- you made a mistake entering your coordinates
- the structure of your system is dominated by
  dispersion interactions.
- you have incomplete k-point sampling

PNS> Dear Ewald
PNS> 
PNS> Thank you for you prompt reply. Here it goes the input (hope enough). 
PNS> All coordinates came from X-ray data (CIF file). I should say that the 
PNS> single point calculation (OPTIMIZE WAVEFUNCTION) has run without problems.

this does not mean a lot. if you plot the charge density,
do you get something reasonable?

BTW: how do you come to expect to get reliable phonons
when your structure is not fully minimized??
please check the literature about what a vibrational
analysis actually does.

cheers,
    axel.

PNS> Thanks again,
PNS> Pedro.
PNS> 
PNS> -----------------------------------
PNS> !  Sample LD01
PNS> !
PNS>  &CPMD
PNS>     OPTIMIZE GEOMETRY XYZ
PNS>     RESTART WAVEFUNCTION COORDINATES LATEST
PNS>     LBFGS
PNS>     CONVERGENCE ADAPT
PNS>      0.02
PNS>     CONVERGENCE ENERGY
PNS>      0.05
PNS>     HESSIAN DISCO
PNS>  &END
PNS>  &DFT
PNS>  FUNCTIONAL BLYP
PNS>  &END
PNS>  &SYSTEM
PNS>    ANGSTROM
PNS>    CELL ABSOLUTE VECTORS
PNS>      9.304 0.0 -6.10537904 0.0 6.6665 0.0 0.0 0.0 10.07482858
PNS>    CUTOFF
PNS>      120
PNS>    POINT GROUP
PNS>      5
PNS>  &END
PNS>  &ATOMS
PNS> *C_SG_BLYP
PNS>  LMAX=D
PNS>  12
PNS> (...)
PNS> *N_SG_BLYP
PNS>  LMAX=D
PNS>  4
PNS> (...)
PNS> *O_SG_BLYP
PNS>  LMAX=D
PNS>  12
PNS> (...)
PNS> *Cl_SG_BLYP
PNS>  LMAX=D
PNS>  4
PNS> (...)
PNS> *H_SG_BLYP
PNS>  LMAX=S
PNS>  32
PNS> (...)
PNS>    CONSTRAINTS
PNS>      FIX SEQUENCE
PNS>        1 32
PNS>    END CONSTRAINTS
PNS>  &END
PNS> -----------------------------------
PNS> 
PNS> Ewald Pauwels escreveu:
PNS> > Dear Pedro,
PNS> >
PNS> > It looks like you might have made some mistakes in defining the 
PNS> > periodicity of your crystal.
PNS> > But please include your inputfile in your post, otherwise people won't 
PNS> > be able to help you.
PNS> >
PNS> > Sincerely,
PNS> > Ewald.
PNS> >
PNS> >
PNS> >
PNS> > Pedro Nuno Simões wrote:
PNS> >> Dear CPMD users.
PNS> >>
PNS> >> I am a beginner, with a basic experience in CPMD. Now, I am trying to 
PNS> >> further expand my capabilities in using the code.
PNS> >> For now I am using CPMD toward the vibrational analysis of organic 
PNS> >> crystalline systems, based on X-ray data. The main problem I am 
PNS> >> facing occurs in geometry optimizations. I was expecting that 
PNS> >> optimization essentially resulted in the correction of protonic 
PNS> >> positions. This was not the case, preventing any reliable comparison 
PNS> >> with experimental vibrational data.
PNS> >>
PNS> >> I have tried several approaches summarized below for a representative 
PNS> >> system.
PNS> >> (1) A cell including 4 complete cations and 4 chlorine anions was 
PNS> >> created based on x-ray information.
PNS> >> Looking to the "trajectory" corresponding to the optimization steps, 
PNS> >> a dramatic structural change was observed, involving 
PNS> >> cleavage/formation of protonic bonds.
PNS> >> (2) I repeated the procedure describe in (1) but by fixing the heavy 
PNS> >> atoms according to a CONSTRAINTS|FIX SEQUENCE specification. The same 
PNS> >> problem was encountered.
PNS> >> (3) I decided to repeat the above scheme by defining the initial 
PNS> >> geometrical parameters directly from CIF data (i.e., without any 
PNS> >> reconstruction of complete fragments within the cell, as attempted in 
PNS> >> (1) and (2). The result was identical to those from the previous cases.
PNS> >> (4) Finally I tried the same as in (3) but using POINT GROUP=5 
PNS> >> (according to X-ray data) instead of POINT GROUP=AUTO in the SYSTEM 
PNS> >> section. In this case, I allowed the optimization to be finished and 
PNS> >> after vibrational analysis has been performed. The results are quite 
PNS> >> doubtful.
PNS> >>
PNS> >> I look forward to any kind of  help.
PNS> >> Thanks in advance,
PNS> >> Pedro.
PNS> >>
PNS> >>   
PNS> >
PNS> >
PNS> 
PNS> 
PNS> 

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