Hi,
I am running an MD simulation for a long linear polymer chain. However, I always get this error: "ERROR on proc 0: Bond atoms 1326 1330 missing on proc 0 at step 456". In order to get it run, I have to reduce the timestep to 0.1fs. But, I want to be able to run the simulation at the timestep of 1.0 fs. I have try to run it with 0.1 fs for 1 million timesteps to get the polymer chain relaxed. Then, I changed the timestep to 1fs, but, I still get the same error. Do you have any idea about this? Thank you. Here is my input script:
####Input Script####
units real
dimension 3
atom_style full
boundary p p p
pair_style lj/cut/coul/cut 15 30
pair_modify mix arithmetic
angle_style harmonic #CHANGE TO HARMONIC
bond_style harmonic
dihedral_style harmonic
improper_style harmonic
read_data poly.dat
velocity all create 500.0 985264378 dist gaussian
neighbor 2.0 bin
neigh_modify delay 10 page 10000000
timestep 1
fix 1 all npt 500.0 500.0 100 xyz 0 0 1000
dump 1 all atom 10000 dump.minimize
dump_modify 1 scale no image yes
thermo 1000
#reset_timestep 0
run 1000000
write_restart restart1
Yin Yani
Can you run it on a single proc for 500 steps and still see the error?
In parallel this typically happens b/c you blow an atom out of the box
and a processor can't find it's bond neighbor for one of its atoms.
Steve
I actually ran it on a single proc, and not in parallel. When I looked at the temperature profile, I found out that the temperature went up so high (to about 1 power of 20) at that timestep, and the two atoms are separated far away which made the bond length bigger than the cutoff value.
Do you have any idea why the temperature can go up to so high value with 1.0fs timestep, but, it ran fine with lower timestep (0.1fs)? Any suggestion about this problem? Thank you.
Yin
Hello Yin
I think the answer is probably quiet trivial. The problem might be that the gradient of your bond potential increases the further you are out of the equlibrium distance. If this is the case your atom could "tunnel" through a lower gradient part of your potential (its tunneling distance is directly proportional to the timestep length) into a region where the gradient of your potential is extremely high. This is typically only the case for distances which are never realised in the real system. Now in the next step it gets an impuls increase according to this extremely high potential gradient and will be able to increase the temperatur on its own by whatever number you can imagine. (Just imagine something like the r^-12 potential. If your atom has speed 0.9 at r=1 and you use timestep 1, then your atom could tunnel to r=0.1 and suddenly it has 10^12 energy (potential energy). If your timestep is only 0.1 it goes to r=0.91 and has only a potential energy of 3.1 and would be stopped before reaching unrealistic distances.)
Btw. we had similar problems as well. Usually this can happen if your input configuration is to far away from equilibrium. It might help to relax the system with small timesteps and as soon as you have dissipated enough energy (using a NVT integrator) you can increase the timesteps.
Hope this helps you a bid
Best regards
Christian
-------- Original-Nachricht --------
A high temperatue and atoms blowing out of the box are all the
same problem - bad dynamics. This could be for a number
of reasons. Bad initial state, bad potential parameters, bad timestep,
etc.
Try to figure out why your atoms are experiencing huge forces.
Steve