The ABINIT Users Mailing List ( CLOSED )

[Audrius Alkauskas <audrius.alkauskas@unibas.ch>]

Hello

I am optimizing the adsorption of a big molecule on a metal surface. 
Since the system is big,
I can play with different geometry optimizers to some extent, but trying 
different parameters will take ages
(usually 1 SCF step is 1 day), therefore I ask people who did such 
calculations, for advice.

Somehow it seems, that quenched molecular dynamics does not treat soft 
and hard modes of the molecule
on the surface at the same footing (I set all the masses to the same 
value).
Soft modes I optimize quickly, but for the hard modes the time step I
chose is presumably too big, and it starts to diverge little by little. 
I don't know whether increasing masses/decreasing time step would help.
In BFGS, the first iteration
should be something like steepest descent, if initial hessian is unit, 
however, already in the first iteration
I get an increase in energy. Does it perform a line minimization or just 
a step in the new direction?

Of course, finally I will find the solution, but it can take long. For 
smaller systems I studied with ABINIT, the conbination
of BFGS in the beginning on quenched MD at the end was working fine.

So, if you have experience about organic molecules on surfaces, please 
let me know!

:)

Audrius