Your symmetry is probably not being broken by numerical noise, which is all
that those tolerances will affect, but rather by the real-space mesh, i.e.
the "eggbox effect."  Increase the mesh cutoff to reduce the symmetry
breaking.  Even better, especially if your system has the symmetry of the
lattice, make the mesh respect the symmetry of your system by e.g.
translating the atoms to put a mesh point on the inversion center of your
system.  Otherwise, you can always explicitly symmetrize your coordinates by
hand after some steps, and continue from there.

David Strubbe
UC Berkeley

On Wed, Jun 24, 2009 at 3:16 PM, Ben Martin <> wrote:

> Hi all,
> Does anyone have any tips for getting better symmetry for geometry
> optimizations/relaxations (MD.typeofrun CG). Right now my relaxations are
> symmetric only to the second or third decimal place, but I would like much
> better symmetry (like to the 6th decimal place).
> I have tried making the DM.tolerance = 10^-7 and DM.EnergyTolerance =
> 10^-7, but that doesn't seem to help much. Does anyone have any tips for
> getting really good symmetry with SIESTA?
> Thanks,
> Ben

Reply via email to