[CPMD-list] Extrapolate WFN

[Axel Kohlmeyer]

On Tue, 3 Jul 2007, Matteo Guglielmi wrote:

ciao matteo!

MG> How does the EXTRAPOLATE WFN keyword work?

the code stores up to N previous wavefunctions and then
does a polynomial extrapolation of the wavefuncion
to derive an improved initial guess for BO-dynamics.

for details, please have a look at the EXTRAPWF() 
subroutine in the file mddiag.F (it is embarrasingly simple).

MG> Does it use the last N wavefunctions to guess the new
MG> one *only* in a BO MD?

yes. where else would you want to use it?

MG> Which value for N would you reccomand?

about 3 to 4. with 4th order extrapolation, the
number of wfopt iteration steps required can be
reduced by a third to half. there is usually not 
much gain in going beyond, except that this may
improve energy conservation (BO-MD is notorious
for losing energy when you use the previous wfn
as initial guess and don't converge _very_ tightly)
and you can get away with a not so tight wavefunction
convergence epsilon (i.e. you benefit even more from
the extrapolation scheme :-) ).

MG> Is it memory demanding?

yes.

cheers,
    axel.

MG> 
MG> Thanks,
MG> MG.
MG> _______________________________________________
MG> CPMD-list mailing list
MG> CPMD-list at cpmd.org
MG> http://cpmd.org/mailman/listinfo/cpmd-list
MG> 

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