I am trying to study a hybrid system with graphene and argon gas.
C-C is AIREBO interaction, Ar-Ar and C-Ar are LJ interaction.
It is a small system. 2000 C atoms form graphene, 15 Argon atoms.
Initially argon atoms randomly distributed in the box and graphene is put on top of the box.
Ar atoms are type 1, Carbon atoms are type 2. I use the following command to specify the hybrid interaction.
But when I do minimization, the potential energy goes to “inf”.
Physically this would not happen. Do I make mistake somewhere?
By the way, before I study the hybrid system, the simulations on individual system all work.
I am trying to study a hybrid system with graphene and argon gas.
C-C is AIREBO interaction, Ar-Ar and C-Ar are LJ interaction.
It is a small system. 2000 C atoms form graphene, 15 Argon atoms.
Initially argon atoms randomly distributed in the box and graphene is put on
top of the box.
Ar atoms are type 1, Carbon atoms are type 2. I use the following command to
specify the hybrid interaction.
But when I do minimization, the potential energy goes to "inf".
Physically this would not happen. Do I make mistake somewhere?
By the way, before I study the hybrid system, the simulations on individual
system all work.
have you checked, that you don't have any overlaps
between the Ar and C atoms?
Thanks! I solved this problem, it is because I set the wrong box size.
Now I got another question. My system includes graphene sheet and argon atoms.
Argon is distributed much more sparsely than graphene. So If I apply parallel computing and my simulation box is divided to small domains according to processor numbers, the graphene sheet which contains most of my atoms is only included in very few simulation domains. The efficiency is pretty low. Do you have any suggestion on this?
Thanks! I solved this problem, it is because I set the wrong box size.
Now I got another question. My system includes graphene sheet and argon atoms.
Argon is distributed much more sparsely than graphene. So If I apply parallel computing and my simulation box is divided to small domains according to processor numbers, the graphene sheet which contains most of my atoms is only included in very few simulation domains. The efficiency is pretty low. Do you have any suggestion on this?
The bigger issue is probably the fact that the computational effort for AIREBO is many times higher than that for an lj potential. I suggest to use the processors command to restrict domain decomposition perpendicular to the graphene layer to a single domain.