Right now I'm trying to deal with the same problem using another order-N code, OpenMX. It does not have many advantages over SIESTA (and in fact, has certain strong disadvantages), but it has that Kerker density mixing scheme in the SCF which allows (in theory) to deal with zero-gap systems. The SCF convergence is really slow, but at least, it's not a divergence! So I think it would be really nice of SIESTA developers to implement something like that in their future releases.
2006/11/8, Oleksandr Voznyy <[EMAIL PROTECTED]>:
> If I want to do some calculations for something adsorbed on the metal > surface, do you mean that it is impossible for me to use the OrderN method? That's right. Although, 100 atoms is not that much and you won't see big difference in calculation time by OrderN vs. Diagonalization.