[Pw_forum] masses and dt in vc-relax

[Jess Kondor]

Hi all,

 I have a question. I am performing 'vc-relax' calculations of the material
that contains quite different atomic masses. In order to speed up
calculations, I tried to find out which timestep (and atomic mass) is more
appropriate. I have chosen a test system (bulk, ~20 atoms) of that material
and tried to vary timestep at fixed atomic mass. For example, I set an
atomic mass for all species to be equal 1.0d0 and found, that dt=70.0 is the
best (fast execution) and further increasing of timestep leads to divergence
of scf cycles. So, next, for the interesting system that is bigger than test
system and has different dimension (surface) I used dt=70 and masses =
1.0However, after several ionic steps the calculations crashed due to
divergence of scf algorithm. I guess this is due to timestep value. I
decreased dt up to 50.0, but it doesn't help. So, I set atomic masses to be
equal their natural values and dt=70. Now the calculations are slow, but
stable. So, the question is, what was wrong with my settings? Does it mean
that I cannot use the settings from the small system and apply them for
bigger system?


-----------------
Regards,
   Jess
-------------- next part --------------
An HTML attachment was scrubbed...
URL: /pipermail/attachments/20070325/70c874fd/attachment.htm