[Pw_forum] Clarification on how to compute elastic constants

[Stefano de Gironcoli]

Dear Ibrahim

     1. in order to calculate the elastic constants you need to start  
from the equilibrium geometry (fully relaxed, both with respect to  
internal degrees of freedom and w.r.t the cell shape) and apply a  
deformation of the cell (say +/- 0.5 or 1.0 %,... you should check  
that you are in the linear regime) and then keeping fixed the cell  
shape relax the internal coordinates. Considering a large enough  
number of independent deformations you should be able to extract the  
elastic constants from the stress-strain relationship.You need to  
perform the internal relaxation because the cell deformation and the  
internal degrees of freedom are in general coupled.

     2. The input of pwscf must contain explicitly all the atoms  
present in the unit cell defined by the fundamental lattice vectors of  
your unit cell. If you use ibrav=0, these are the ones you define in  
the CELL_PARAMETERS card...
      as for the example of ZnSe it contains 8 atoms if your unit cell  
is the simple cubic one but only 2 if the fcc fundamental lattice  
vectors are considered.

      3. As for the full structural optimization... vc-relax should  
work well (check carefully that stress calculation in converged).  For  
simple systems, such as SC, FCC, or BCC, a lattice parameter scan + an  
Equation of State fit (ev.x) should also work fine.

      hope this helps,

stefano


Quoting bamidele ibrahim <bamideleibrahim at yahoo.com>:
>  Dear all,
> Please, i need some clarification on the following  issues on how to  
> calculate elastic constants,
> 1. When performing the calculation, is it correct to set  
> 'calculation=scf' or 'relax'? For either of the two cases, can   
> anybody in the house give me an explanation on it.
> 2. since we need to set ibrav=0, and define explicitly the primitive  
> vectors for any crystal we are dealing with then, is it necessary to  
> define all the atoms in the primitive cell of the crystal  
> explicitly?  for example ZnSe which has 8 atomic positions if  
> ibrav=0 and this in the cubic B3 phase.
> 3. Do i need to use cell dynamics or ion dynamics for the  
> optimization of the cell?
>
> I will be delighted if anybody can give me an explanation on this issue.
>
> Thanks
>
>  
> Adetunji Bamidele Ibrahim(PhD Student)
> Department of physics,University of Agriculture,
> Abeokuta, Ogun State,Nigeria.