[lammps-users] NVE energy conservation

Hi, I found that the energy conservation is very good in 10 ns if I run simulations with boundary conditions p p m on a single layer of atoms with NVE ensemble, but the energy conservation is very poor if I use boundary conditions p m m, in 100 ps the temperature increases about 2 K, and it gets worse and worse in longer simulations times.
What is the reason, and how can I improve the energy conservation in the latter case?
I used the Tersoff potential.

Thanks.
Dong

Hi, I found that the energy conservation is very good in 10 ns if I run
simulations with boundary conditions p p m on a single layer of atoms with
NVE ensemble, but the energy conservation is very poor if I use boundary
conditions p m m, in 100 ps the temperature increases about 2 K, and it gets
worse and worse in longer simulations times.

how the hell should somebody know without seeing what kind of system
you are running.

What is the reason, and how can I improve the energy conservation in the
latter case?

one thing to consider is the following: with a change in the boundary
conditions you change your system and the resulting system may no
longer be in equilibrium.

but this is pure speculation based on some vague guess of what
might be happening.

cheers,
    axel.

The system is graphene, a single layer of carbon atoms.
I run NPT first, then NVE with either p p m or p m m boundary conditions.
The system has changed, but in either case I have equilibrated the system with NPT run.
Any more ideas?

Thanks.
Dong

Did you look at your system visually to find out what is happening? Is it drifting or just vibrating around there equilibrium positions?

The edge atoms are more loosely bound if you do not use periodic BCs.
Thus they are effectively 'lighter' than atoms in the middle and their
velocities are higher. You need to reduce the time step to get around
that problem.

Zhun-Yong