Actually, I changed the Mesh Cut-off to 500 Ry. There are three non-trival 
frequences. two are near 1033.9, and one is near 3372. And there is no decrease 
trend when I increase the Mesh-cut off.

Quoting [email protected]:

>> I am trying to calculate the vibration mode of H2 molecure. I put H2
>> into a unit cell of a lattice, and then make the lattice constant equal
>> to 10 Ang ( I think this will prevent the interaction of different H2
>> molecure in different unit cell). However, after following the way (
>> using fcbuild to make FC.fdf and using siesta to calculate FC constant
>> and then analyze using vibrator), I get 6 vibration mode, while 3 are
>> trivial, but the other three are not). As H2 is a linear molecure, I
>> think the number of vibration mode should be 1. I am not sure what's
>> wrong with my calculation.
>
> Probably nothing, but Siesta doesn't know about the symmetry
> and produces always 3N frequencies. Of which three must correspond to
> uniform translations and be more or less zero, either for crystal or
> molecule, and moreover, in general for a molecule, three rotations,
> whose frequencies should also be nearly
> zero. In case of diatomic molecule however,
> one rotation (along the molecule axis) leaves the molecule invariant,
> so it should not be counted among these "trivial" modes, and you
> stay instead with one "genuine" vibration mode.
> It is relatively easy to get zero
> frequency of three translation modes (setting MeshCutoff high enough);
> the rotational modes are more tricky - I think, it is simply more
> numerically demanding. Probably you got them not close enough to zero,
> due to unsufficient isotropy of "Siesta space" (mesh, basis etc.).
>
> Best regards
>
> Andrei Postnikov
>
>

Attachment: h2-siesta.fdf
Description: application/vnd.fdf

Attachment: h2.fdf
Description: application/vnd.fdf

Attachment: FC.fdf
Description: application/vnd.fdf

Attachment: H2.bands
Description: Binary data

Attachment: H2.vectors
Description: Binary data

Responder a