Eotal shows a slightly fluctuates under the NVE ensemble

LAMMPS LAMMPS Mailing List Mirror
1

Dear
I use a rigid projectile to impact the monolayer h-BN membrane. The projectile is consist of some atoms.
The system has been fully balenced in nvt and npt. During the impact, I find the etotal is not conserved.
More specifically, the system is initially conserved. Once the atomic bonds of the membrane are broken (step>3520), the etotal began to rise
The data are as follows:

step etotal
3280 -62615.53
3320 -62615.53
3360 -62620.531
3400 -62626.532
3440 -62633.53
3480 -62643.532
3520 -62645.531
… …
11160 -62395.255
11200 -62395.255
11240 -62395.255
11280 -62395.255
11320 -62395.255
11360 -62395.256

Although the fluctuation amplitude is not large, it still does not meet my analysis requirements.
I have reduced the timestep but it didn’t work. If the amplitude can be controlled at a low level, such as <50, that’s will be great.
What should I do?
Here is my manuscript:

clear
read_restart h-BN.restart
neighbor 3.0 bin
neigh_modify every 1 delay 0 check yes one 10000

reset_timestep 0
timestep 0.0001

pair_style hybrid/overlay tersoff lj/cut 5
pair_coeff * * tersoff BNC.tersoff N B NULL
pair_coeff 1*2 3 lj/cut 1.0 2.5
pair_coeff 3 3 none
thermo 40
thermo_style custom step etotal

velocity boundary set 0.0 0.0 0.0
fix F1 boundary setforce 0.0 0.0 0.0
fix F2 membrane nve
velocity projectile set 0.0 0.0 -80 units box # 8km/s
fix F3 projectile rigid/nve molecule

run 30000

2

Dear
I use a rigid projectile to impact the monolayer h-BN membrane. The projectile is consist of some atoms.
The system has been fully balenced in nvt and npt. During the impact, I find the etotal is not conserved.
More specifically, the system is initially conserved. Once the atomic bonds of the membrane are broken (step>3520), the etotal began to rise

you are using *very* unusual parameters for your Lennard-Jones
interactions. the interaction strength (epsilon) is extremely large
for metal units and at the same time you have a very small effective
radius (sigma) and an extremely small cutoff and no smoothing or
long-range treatment of the attractive lennard-jones term.

have you checked, whether the tersoff potential you are using is
suitable for modeling bond breaking. with BN i would expect the
creation of polar surfaces, which i would not expect to be modeled
well with tersoff style potentials. i would expect that this would
require some form of charge equilibration procedure to be included in
the calculation.

axel.