Questions about initial velocity for annealing O2 molecules

Dear LAMMPS Users,

I am interested in simulation of annealing O2 molecules with NVT ensemble at 300 K.
The Si-O Tersoff potential was applied in this simulation.
The main input script and simluation input files is provided below in this post.

If an initial velocity value is not set for the oxygen molecules, they remain stationary within the simulation box and only vibrate in place.
Conversely, when an initial velocity value is assigned using the ‘velocity’ command, the oxygen molecules begin to aggregate.

My initial inquiry pertains to the necessity of specifying an initial velocity value.
Additionally, I’m wondering whether there may be any issues with the Tersoff potential parameterization I employed, particulary in regard to the observed clumping of oxygen molecules.
The employed Tersoff potential is the well established parameterizaton proposed by S. Mumetoh et al.

Thank you

[without initial velocity]

[with intitial velocity]

[main input file]
##----------------------------------Initialization Settings---------------------------------------
units metal
dimension 3
boundary p p p
atom_style atomic

##-------------------------------System Definition-------------------------------
variable Lx equal 15.361188
variable Ly equal 15.361188
variable Lz equal 31.0
variable vtemp equal 300

region simulation_box block 0 {Lx} 0 {Ly} 0 ${Lz}
create_box 1 simulation_box
read_data O2.dat add merge

mass 1 15.999400
##---------------------------Potential Settings----------------------------------
pair_style tersoff
pair_coeff * * SiO.tersoff O

timestep 0.001 #1fs (default dt = 1ps)
##-------------------------Simulation Execution---------------------------------
thermo 1000
velocity all create ${vtemp} 231144
dump 2 all custom 100 md.lammpstrj id type x y z

fix 1 all nvt temp {vtemp} {vtemp} $(100.0*dt)
run 100000

write_restart end.rset
write_data end.dat

[simulation files]
in.O2 (1.3 KB)
md.lammpstrj (3.9 MB)
O2.dat (2.8 KB)
out.O2 (9.5 KB)
SiO.tersoff (1.8 KB)

This is a VERY bad choice. This potential has been parameterized for bulk SiO2 systems (crystalline and glass) and nothing else.

Thanks for the reply,

I’ve missing something important.
Additionally, apart from this SiO potential employed in this simulation, I was wondering if this means that the tersoff potentials exclusively designed for bulk materials.

Thank you

Not all of them. But particularly those for polar components would be. Since the Tersoff formulation does not include explicit charges and specifically oxides typically require charge equilibration to get meaningful representations of surfaces, edges and corners or tips. Potentials with carbon and elementary silicon have been used for nano particles or nano-machining etc. There is one example of the latter somewhere on the LAMMPS home page, IIRC.

For the SiO.tersoff potential in question, there have been documented cases of its failure for nano-particles or surfaces in the LAMMPS mailing list archives.