The ABINIT Users Mailing List ( CLOSED )

[Konstantin Rushchanskii <Konstantin.Rushchanskii@physik.uni-regensburg.de>]

Dear all,

exploring fileSrc_9drive/brdmin.f I have found, that the coordinates for 
the next iteration
are obtained from the old coordinates, hessian and forces in the form:

!   New atomic cartesian coordinates are obtained from vin, hessin and vout
   do idim=1,ndim
    vin(:)=vin(:)-hessin(:,idim)*vout(idim)
   enddo

I would like to ask, would'nt it be more flexible if a factor (similar 
to dtset%strprecon,
but less than unity) is added  to tune an amplitude of coordinates 
changing,
especially for the systems that have more that one local minima?

Say, in the form
    vin(:)=vin(:)-hessin(:,idim)*vout(idim)*dtset%broydenmix

I have a trouble with finding the equilibrium configuration of  
ferroelectric phase starting from the experimental data,
because my system is falling down to the equilibrium paraelectric phase 
configuration with higher symmetry.
I suppose, the reason is a big amplitude of the forces in the first 
Broyden steps, and the application of  the mentioned
factor could control the minimisation process.

Sincerely,
K.R.


--
K.Rushchanskii, Dr.
Theoretische Physik, Universität Regensburg
93040 Regensburg Germany
tel: +49 (0) 941-943-2466  fax: +49 (0) 941-943-4382