Unfortunately, I don't know much about SHAKE, so I am going to have to
forward this to the LAMMPS mailing list. Does anybody have some ideas
on how to track down this problem?
hho.txt (282 Bytes)
in.parameter (1.23 KB)
data.water (486 KB)
Unfortunately, I don't know much about SHAKE, so I am going to have to
forward this to the LAMMPS mailing list. Does anybody have some ideas
on how to track down this problem?
if the SHAKE determinant is zero, that means, the constraint equation
is degenerate, e.g. in the case of water it can happen, if all atoms
would be on a line after the unconstrained position update.
axel.
Yongbiao,
There are just too many possible problems with your script. It is
impossible to say which one might be causing the water molecules to
mysteriously become linear (a very bad sign). Try to isolate the cause
by eliminating things that don't change the behavior e.g. fix
momentum, fix gcmc. Also, try to see what else is wrong with the
simulation. Things are probably going bad long before the fatal error
message.
Aidan
hi aidan,
there are a bunch of things to be considered on top of this, that are
evident from reading the documentation.
to have newly inserted molecules "shake"-n, the molecule template
needs to be set up for this *and* fix gcmc instructed to inform fix
shake about newly added or removed shake-clusters. also, one has to
take correctly care of the energy from exclusions (which need to be
specified in the template as well). given that this is all rather new
and - at best - lightly tested functionality, there is also the chance
of some bug in the code here and there. i would not recommend anybody
to work on this unless this person has already experience on GCMC from
other codes, has a good understanding of the inner workings of LAMMPS
and sufficient experience in C++ programming and (parallel) debugging.
i am fairly certain, that there are some subtle issues to consider in
SHAKE, that may work OK for diatomics, but not for water.
axel.