[CPMD-list] Molecular crystals !!

[Axel Kohlmeyer]

On Mon, 10 Dec 2007, Belabbas Imad wrote:

BI> Dear CPMD users,

dear imad,

BI> I am currently trying to calculate the lattice parameters of a 
BI> monoclinic molecular crystal : Acetonitrile (C2NH3). For that I am 
BI> relaxing the system for several values of the lattice parameters 
BI> (the classical procedure !!).

BI> * Is it possible to relax the directly lattice parameters within 
BI>   CPMD ? (doing variable cell relaxation)

technically yes. you could do variable cell NPT dynamics with 
annealing on coordinates and cell dimensions (and electrons), 
but it is not necessarily accurate and rather difficult to get 
right (choice of fictious mass for electrons and cell, cutoff, 
time step, 3 annealing factors etc. etc.).

BI> During my relaxations, I have encountered to following message:
BI> 
BI> ODIIS I Insufficient progress for K= 2; reset !
BI> 
BI> and one single geometry relaxation requires about 30 SCF cycles !!

convergence for complex wavefunctions at different k-points
is not always as fast as for the (real) wavefunction at gamma.
30 cycles is not too many.

BI> 
BI> * Is there a way to improve this ?

BI> I have tried to use CG for the WFN relaxation instead of ODIIS but 
BI> it seems that the method has not been implemented yet for periodic 
BI> systems . (I had the following message: PROGRAM STOPS IN SUBROUTINE 
BI> k_pcgrad| Line Search not implemented yet )

conjugate gradient works. only the line search does not.
simply omit the additional keyword MINIMIZE that turns on
the line search and you'll get the regular conjugate gradient
also for k-points.

BI> * is the value of 5x1e-04 for the geometry convergence is reasonable 
BI>   !!!??

that depends a lot on how accurate DFT can describe
your system at all. there is no point in optimizing
something tightly that has a large error bar to begin with...

BTW: your LMAX=P invalidates LOC=D and will turn it into
  LOC=P. this is how it _should_ be for the pseudopotentials
  from the CPMD psp libraries, but i suggest you change the
  input to not get fooled by it.

cheers,
   axel.
BI> 
BI> I am attaching the input file that I am using !!
BI> 
BI> Best regards,
BI> 
BI> Imad
BI> 
BI> 
BI> 
BI>              
BI> ---------------------------------
BI>  Ne gardez plus qu'une seule adresse mail ! Copiez vos mails vers Yahoo! Mail 

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