[Pw_forum] The cell blowing up in MD simulation

[Paolo Giannozzi]

On Thu, 2012-09-13 at 15:36 -0700, Tian Lan wrote:

> But it seems that the energies that both codes obtained agree well. I
> pasted the results in the following. The first one is the result from
> vc-relax operated on one unit cell, the total energy is -358.1 Ry,
> this is very close to my original setup with scf calculation. The
> second one is from cp.x operated on 2 by 2 by 2 supercell, with fixed
> atom positions(ion and cell_dynamics are set to 'none'), the energy is
> -1410 au==-2820 Ry. I don't know whether this is a big difference

the energy is -1431.80272 Ha or so, i.e. -1431.80272/8 Ha per cell,
i.e. 357.95Ry. Close (358.097) but not very close. Note however that 
you need to compare equivalent results, i.e. same cell, same k-points,
or a cell and a supercell but with equivalent k-points. Also note 
that the treatment of the augmentation charge via the "small boxes"
requires some testing in order to find a suitable value for input 
variables nr1b, nr2b, nr3b. The resulting pressure may also change 
quite a bit.

> I don't quite get your point about the default positions of atoms in
> pw and cp. In both codes, I used 'crystal'

perfect: it is of no concern for you. If you do not specify anything,
cp will assume borh radii and pw lattice parameters, or something
like this, as explained in the documentation

P.
-- 
Paolo Giannozzi, IOM-Democritos and University of Udine, Italy