Error: Lost atoms: Original 38496 current 38493 (../thermo.cpp:389).

Dear Lammps users

I want to keep my system in room temperature 300k at first. Then I
want to apply velocity on graphite which will move to Cu Substrate
with certain velocity.
To maintain the temperature I used gaussian distribution. But found
after some times it is appearing some atoms lost from the simulation
box and showing
error: Lost atoms: Original 38496 current 38493 (../thermo.cpp:389).
To over come that problem priviously I enlarged my simulation box in
every side much
more and found bottom layer atoms are flying away to go lower energy
state, which has been attached in screen capture 6. Finally I tried do
it by enlarging very few cells
only in +ve X and Y directions. But found same error, which mentioned
in capture4. Is it for gaussian distribution command? How to solve
this problem and should I
use another command to control this temperature? I am really
appriating an way of solution. Here I am writting my input file...

# LAMMPS Input File

# ---------- Initialize Simulation ---------------------
units metal
dimension 3
boundary p p f
atom_style full

# ---------- Create Atomistic Structure ---------------------
read_data system1.LAMMPS05
group substrate type 1
group graphite type 2
region wall block INF INF INF INF -8.1500 -7.1529 units box
group bottom region wall
region middle block INF INF INF INF -7.1529 7.287090 units box
group stable region middle
region upper block INF INF INF INF 7.287090 108.91000 units box
group top region upper
region upper2 block 0.173601 99.448601 -0.966524 98.308502 28.947100
30.752100 units box
group surface region upper2
#fix 1 stable setforce 0.0 0.0 0.0

# ---------- Define Interatomic Potential ---------------------
pair_style hybrid eam lj/cut 8.0 airebo 4.0 0 0
pair_coeff * * airebo CH.airebo NULL C
pair_coeff 1 1 eam Cu_u3.eam
pair_coeff 1 2 lj/cut 0.1 3.225
neighbor 2.0 bin
neigh_modify delay 10 check yes

# ---------- Run Minimization ---------------------
timestep 0.001
velocity all create 300.0 3213112 mom yes rot yes dist gaussian
fix relax all nvt temp 300 300 0.01
thermo 100
thermo_style custom step temp ke etotal
dump 1 all xyz 500

restart 100000 deposite.restart
run 100000




You are using a “f” = fixed boundary in z. Which
means that if atoms cross that boundary they
will be lost. If they shouldn’t cross you need
to figure out why they are, and fix it.

If it’s OK to lose those few atoms, you can

use thermo_modify lost to tell LAMMPS to
ignore lost atoms and continue.


Steve just beat me to it.
From the names of your regions I assume this is because you want to put a wall there, but I don’t see any fix for that?