[Pw_forum] about structural optimization

[Stefano de Gironcoli]

sun jason wrote

>
> I have some questions about structural optimization
> 1) can I just optimize the lattice constants but not change the atomic
> fractional coordinates? 

Why do you want to do that ? If positions are fixed by symmetry (like in
diamond structure for instance) they
will not move even if they are allowed to. If they are not fixed by
symmetry then it is incorrect to fix them.
Anyhow, in the input file there is the possibility to lock certain
atomic positions by specifying three additional numbers
per atoms as in the following example (check the Doc/INPUT_PW file for
an explanation)
Si 0.00 0.00 0.00 0 0 0
Si 0.25 0.25 0.25 0 0 0
It should work also in combination with cell-dinamics (the force on the
locked atoms are zeroed before entering MD routines)
Another option is obviously to manually change the lattice parameter w/o
changing the atomic positions (in crystal coordinates)

>
> 2) does ion_dynamics = 'bfgs' can be use to vc-relax?
> I can run a vc-relax calculation when I put ion_dynamics = 'damp', and
> it says does not support 'bfgs' when I put ion_dynamics='bfgs'

No. vc-relax and vc-md employ routines contributed by Renata Wenzcovitch
and coworkers and do not support bfgs minimization.

>
> 3) I run an example to calculate the effective charges of cubic BN, to
> my knowledge, the absolute value of the effective charge for cation
> and anion should be equal in this case. but I get 1.759 and -1.736 for
> B and N atom, respectively. why? or just because I have not relax the
> cell?
> any help would be appreciate.

effective charges should satisfy the so called acustic sum rule ( \sum_i
Z_i = 0) .
However in practice this is never realized due to incomplete BZ
sampling, other technical limitation like not-optimized structure,
finite cutoff or FFT grids are not relevant for ASR violation.
One can easily IMPOSE the ASR obtetaining accurate results even with
modest BZ sampling.
This isssue is discussed in P.Giannozzi et al, PRB 43, 7231 (1991) .

Stefano de Gironcoli