The ABINIT Users Mailing List ( CLOSED )

[Xavier Gonze <gonze@pcpm.ucl.ac.be>]

Dear taoj0003,

Although one might envision converging together all lattice parameters
and internal atomic positions (using optcell=2) in one shot, there are
strong non-linear coupling effect between lattice parameters and  
atomic positions.
So, in practice, it is better to optimize first the atomic positions  
in fixed cell parameters,
because this is often associated with a large energy change, while  
being fairly
harmonic. When the atomic positions are OK, feed them in a optcell=2  
run,
and do the full optimization of everything (which might still not be  
easy).

Unless you definitely want to keep the ratio between a, b, c fixed  
(and the angles if they
are not fixed by symmetry), you must allow all of them  to relax. The  
theoretical
and experimental ratio will not be the same (as all other parameters).

Xavier

On 30 Apr 2007, at 05:35, taoj0003@ntu.edu.sg wrote:

> when using abinit(maybe also some other software) to optimize the  
> structure, a problem really puzzled me-what is the best procedure?
>
> let me make this more clearly. To optimize the structure, here is  
> one procedure:
>
> First, if there are more than one atom in the cell, we need feagure  
> out all of the optimized coordinations. Abinit make this part easy  
> to do thanks to the "ionmov".
>
> Then we need to optimize lattice constants.But how? Just keep their  
> experimental ratio and optimize the volume or adjust every values  
> (a, b ,and c) near the experiment values until it reached the  
> lowest total energy? If you do it the first way, I may ask, is the  
> experiemnt ratio correct for using in the simulation?( because,  
> simulation always deviates from experiment somehow).
>
> The second part really make me crazy. There are also some cross  
> section effect. And even should the seccond part go first?