Dear Roberto,

   Maybe you could start with reading "Systematic generation of
finite-range atomic basis sets for linear-scaling calculations". That
paper could be downloaded on siesta website. Another good paper about
basis set is "Numerical atomic basis orbitals from H to Kr, PHYSICAL
REVIEW B 69, 195113 2004", they use a different scheme to generating
basis set.

   For tools,  Util\Optimizer\ has two tools to get an optimized basis set.

  Generally, to optimize basis set, the idea is to get a lower energy
(this is also true for optimizing pseudopotential ). However, for
using atomic basis set, it is not as variational as plane wave. Tuning
parameters may seem arbitrary even though they have physical meaning.
The over complete or under complete basis set may have some kind of
influence in some extent in the computing result, however, I didn't
find a detailed report about direct relation between them. I mean the
relation between incomplete of basis set and the computing result.
Only plane wave basis set is believed to be progressively complete
when one increase energy cutoff.  I'm not an expert on basis set,
please correct me if anything wrong.

  In my experience, even without any optimized basis set, the result
should look good, I means the tendency, like 'Energy-lattice
parameter' curve. Maybe you could compute the lattice constant of
silicon for  some experience.

Best Wishes,
lukuilin

On Thu, Oct 28, 2010 at 10:53 PM, Roberto Guerra <[email protected]> wrote:
> Dear lukuilin,
>
> I'm surely less familiar than you with siesta bases (I usually don't
> specify any basis parameter). Such a sensitivity to the basis definition is
> getting me scared. Do you know if is there any procedure assuring that the
> basis is good? I'm mostly working with silicon and oxygen.
>
>
> Best,
> Roberto
>

Responder a