[CPMD-list] Bad results during the heating phase of a md run

[Juerg Hutter]

Hi

there seems to be several problems with your simulation.
Do you start the simulation from optimized wavefunctions?
In your first simulation you have very large initial forces
(up to 15!). This will cause a large initial step and
the wavefunction (even when converged) will not be able
to follow.

You should first relax the geometry either using
Geometry optimization or MD with annealing and then
start an MD and test for the optimal EMASS and
timestep.

regards

Juerg Hutter

----------------------------------------------------------
Juerg Hutter                   Phone : ++41 44 635 4491
Physical Chemistry Institute   FAX   : ++41 44 635 6838
University of Zurich           E-mail: hutter at pci.unizh.ch
Winterthurerstrasse 190
CH-8057 Zurich, Switzerland
----------------------------------------------------------


On Tue, 23 Aug 2005, Reinaldo Pis Diez wrote:

> 	Dear folks,
>
> 		I'm trying to reproduce some older results on sodium clusters
> (Röthlisberger and Andreoni, JCP 94 (1991) 8129) following a different
> strategy. I want to let the clusters evolve at a given temperature for
> a long period saving the coordinates every several ps and then
> optimize each previously saved geometry.
> 		I'm using version 3.9.2 of CPMD, Goedecker pseudos for sodium with
> Zion = 9 and PBE xc functional. I've made some previous tests with the
> sodium dimer and found that 60 Ry for PW cutoff are ok to reproduce
> the bond distance, for example. Then, I turn to the sodium hexamer. I
> use a simulated annealing procedure to optimize the wavefunction
> (na6_wfn.inp and na6_wfn.out files) and it seems to be alright. Then,
> the hexamer is allowed to heat up to 600 K. The lattice of the (cubic)
> cell was 15 A and the PW cutoff was 60 Ry as said above. The md run
> diverges from the second time step. I increased the lattice cell and
> the PW cutoff to 25 A and 70 Ry, respectively, to see where the
> problem is but I got the same results (na6_md_ini.inp and
> na6_md_ini.out files). I can see from the output that EKINC is going
> large at the very beginning of the run but I don't know if this could
> be a clue and how to handle it to  get the simulation running. Any
> help will be very appreciated. The input and output files are appended.
> 		Sorry for the long mail but I cannot figure out how to explain my
> problem in few words.
> 		Thanks in advance.
> 		Regards,
>
> 						Reinaldo
>