Reax/c & neigh_modify exclude

Dear Lammps user and Developers

I tired to simulate the crack by using the Reax/c potential and “neigh_modify exclude group” command. the simulation bpx have 7000 atoms and 150 atoms in each crack group.

during simulation I get the following error.
without crack there is not proble. please advise me why this happend?
WARNING: Fix qeq/reax CG convergence failed after 200 iterations at 16 step (…/fix_qeq_reax.cpp:639)
.
.
.
WARNING: Fix qeq/reax CG convergence failed after 200 iterations at 50 step (…/fix_qeq_reax.cpp:639)
step49-bondchk failed: i=1182 end(i)=36744 str(i+1)=36739
application called MPI_Abort(MPI_COMM_WORLD, -14) - process 1
step49-bondchk failed: i=9159 end(i)=301088 str(i+1)=301086
application called MPI_Abort(MPI_COMM_WORLD, -14) - process 0

Dear Lammps user and Developers

I tired to simulate the crack by using the Reax/c potential and “neigh_modify exclude group” command. the simulation bpx have 7000 atoms and 150 atoms in each crack group.

during simulation I get the following error.
without crack there is not proble. please advise me why this happend?

Fix qeq/reax failed to converge when calculating equilibrium charges, some atoms acquire unstable charge values (usually big), which then resulted in huge Coulombic forces. Then atoms move too fast, then the bond list break down.

R.

1 Like

Dear Ray

Thanks for your reply. is there any suggestion to avoid of this error?

Thanks

Dear Ray
Thanks for your reply. is there any suggestion to avoid of this error?

not sure if this helps, but i'd suspect that this kind of neighborlist
manipulation, may not give consistent results with manybody
potentials. have you tried just punching out some atoms to induce the
crack?

axel.

Thanks Axel. if you mean is delete the atom, NO I didn’t do that. but I think if I deleted a group of atoms to create crack, the atoms there will bounded again during equilibration or production.
am I wrong or not?
Thanks

Thanks Axel. if you mean is delete the atom, NO I didn't do that. but I
think if I deleted a group of atoms to create crack, the atoms there will
bounded again during equilibration or production.
am I wrong or not?

if i knew the result of simulations ahead of time, what would be the
purpose of simulations?

axel.

Dear Axel
I have done this simulation by deleting 3 rows of atoms. the crack length was about 20 angstrom.
(maybe I should delete more rows of atoms). as I think the atoms start to bound to each other.
I think I should define there is no potential between this two groups. but don’t know how do this if I couldn’t use “neigh_modify exclude group”.
is there any suggestion
Thanks

Dear Hadi,

why do you think there should be no interaction between crack boundaries? Are you sure that if you artificially switch off these interactions you will obtain physically meaningful result? Maybe if bonding occurs again under given conditions - then it should occur under these conditions, and you should seek for more appropriate ones? Say, try to vary external force or initial crack size.

I think that turning off all the forces inside the crack when you simulate this crack is not the best idea.

Regards,
Oleg Sergeev,
VNIIA.

17.06.2013, 14:05, "Hadi Sedaghat" <[email protected]>:

Dear Axel
I have done this simulation by deleting 3 rows of atoms. the crack length
was about 20 angstrom.
(maybe I should delete more rows of atoms). as I think the atoms start to
bound to each other.
I think I should define there is no potential between this two groups. but
don't know how do this if I couldn't use "neigh_modify exclude group".
is there any suggestion

my suggestion is to not simply blindly follow what other people are
saying or doing or have done, but rather understand what the problems
are and solve them. so far, i only see you being interested in
suppressing symptoms. i've made a suggestion for that, but if that
doesn't work for you, then it doesn't work. it is not that a
suggestion from somebody on this list is a fundamental law that must
not be violated. but at the same time, you either have to provide more
information and make a more details and in-depth analysis of what is
*really* causing the problem (is it perhaps something completely
different, that is only exposed by running a crack simulation?), or
figure it out on your own.
my patience with "just tell me what to do"-people is rather limited.

axel.

All that neigh_modify group exclude will do is delete
certain pairs from the pairwise neighbor list. For a manybody
potential like ReaxFF there could still be other ways atoms
across the crack interact, and more generally, deleting
some pairs from the pairlist could lead to strange behavior.
I would not suggest doing a crack that way with ReaxFF.

Steve

Additionally, with fix qeq/reax it should be fine if an atom does not see other nearby atoms that are excluded from the neighbor list while equilibrating charges. It just equilibrates its charge as if those excluded atoms are outside the cutoff. However, each group that fix qeq/reax operates on must be charge neutral. If not, the fix will make it neutral, and this may cause atoms to acquire weird charges and may be one of the reasons for the convergence failures.

Ray