I am trying to simulate large MoS2 sheet in water. Sheet is held at its place by being excluded from time integration in npt.
I am running it on Xeon PHI using user-intel package hence cannot give neigh_exclude. So to avoid energy contributions from motionless atoms i decided to delete their bonds and set lj paramenetrs to zero.
However when i give:
pair_coeff 5 5 0.0 2.551 ## Mo Mo interactions
pair_coeff 8 8 0.0 3.5 ## S S interactions
simulation crashes with missing atoms error. I can see in trajectory file that water molecules come too close to S the just shoots away. Why is pair coeff between 8 8 affecting pair coeff of other pairs?
pair coeff for all atoms are provided in lammps data file under "Pair Coeffs" section.
LAMMPS version: 12 oct 2016
Is this the intended behavior? If yes what could be correct approach?
Also when I try to maintain constant pressure, due to anisotropic nature of simulation cell, pressure is way off (3000 atm). To maintain it i decided to manually increase volume in fix nvt to get PV curve. However when i do increase volume, pressure starts to come down but thermostat is overcooling. I would love if someone more experienced can shed any light on how they maintain pressure constant for overtly complicated systems
I am trying to simulate large MoS2 sheet in water. Sheet is held at its place by being excluded from time integration in npt.
I am running it on Xeon PHI using user-intel package hence cannot give neigh_exclude. So to avoid energy contributions from motionless atoms i decided to delete their bonds and set lj paramenetrs to zero.
However when i give:
pair_coeff 5 5 0.0 2.551 ## Mo Mo interactions
pair_coeff 8 8 0.0 3.5 ## S S interactions
simulation crashes with missing atoms error. I can see in trajectory file that water molecules come too close to S the just shoots away. Why is pair coeff between 8 8 affecting pair coeff of other pairs?
pair coeff for all atoms are provided in lammps data file under "Pair Coeffs" section.
LAMMPS version: 12 oct 2016
Is this the intended behavior?
yes.
you are only providing a subset of the parameters for all
interactions. the remaining (mixed) parameters are inferred from
mixing rules. when you set epsilon for Mo-Mo and S-S to zero, you are
also changing the interactions for all *mixed* pairs that involve Mo
or S. in the case of Mo-S this is desired, but for all interactions
with Mo or S and all other atom types, this is wrong.
If yes what could be correct approach?
you need to explicitly provide all mixed terms. you can issue the
"write_coeff" command to output the complete set of pairwise
parameters immediately after reading the data file. once you have this
file, you can delete the pair coeff section in the data file and
instead use the list of explicit pair_coeff commands instead and
selectively set only Mo-Mo, S-S and Mo-S to zero.
axel.