I have been trying to write an input file for to simulate the shape transition of a nanoparticle of silicon using multiple potentials to obtain the best structure, which I will compare with a simulation results. However, I have encountered an issue with the structure of the nanoparticle being distorted from the beginning when using the fix ID nve command, and I don't know how to fix it.

"an example of 13 particles using lammps"

I've tried to change the fix nve command with the fix nvt command or just leaving the velocity command running alone, but the resulting structure is not as expected when compared to the results.

[tag:Density-functional-theory] results

Here is the input file that i have been using:

variable        dt equal 0.001
units           metal

variable t equal 800
variable t_start equal 300
variable t_final equal 1

# ----------------  Loop start -------------------------------------
variable i loop 12 26
variable cluster equal $i
variable epair_cluster equal epair/v_cluster
variable num_steps equal (${t}-${t_final})*1000

#-----------------------ATOM DEFINITION----------------------------
read_data       data/si_${i}.data 
mass            1 28.085  ##Silicon

#------------------------FORCE FIELDS------------------------------
#pair_style      sw
#pair_coeff      * * Si.sw Si

#pair_style edip
#pair_coeff * * Si.edip Si

pair_style tersoff
pair_coeff * * Si.tersoff Si

#pair_style     tersoff/mod/c 
#pair_coeff * * Si.tersoff.modc Si

neighbor        2 bin
neigh_modify    every 10 delay 0 check no

thermo_style custom step temp v_epair_cluster etotal v_cluster
thermo          1000

#dump            0 all xyz 100 min/si_min_$i.xyz
#dump_modify     0 flush yes
#dump_modify     0 sort id element Si
#minimize 1.0e-4 1.0e-6 100 1000
#undump          0

reset_timestep  0
timestep        $(dt)

#------------------------------ Setting the velocities  -------------------------
fix         1 all nve 
velocity        all create  ${t_start} 4928459 rot yes dist gaussian

dump            5 all xyz 1000 si_$i_$t.xyz
dump_modify     5 append yes

run     250000

unfix       1
#------------------------------- annealing -------------------------
fix             1 all nvt temp ${t_start} ${t} 0.1

dump            1 all xyz 1000 si_$i_$t.xyz
dump_modify     1 append yes

run             ${num_steps}

#------------------------------- Relaxation  -------------------------
fix             2 all nvt temp  ${t} ${t} 0.1

dump            2 all xyz 1000 si_$i_$t.xyz
dump_modify     2 append yes

run             250000

#-------------------------------  couling -------------------------
fix             4 all temp/berendsen ${t} ${t_final} 0.1

dump           3 all xyz 1000 si_$i_$t.xyz
dump_modify    3 append yes

run             ${num_steps}

#------------------------------- Relaxation -----------------------
fix     5 all nvt temp ${t_final} ${t_final} 0.1

dump           4 all xyz 1000 si_$i_$t.xyz
dump_modify    4 append yes

run            250000

 next i
 jump annealing_si_loop.lmp

print         "Simulation complete"
  • 4
    $\begingroup$ Your description of the problem is quite vague. What exactly are the DFT results in your case, and what are you getting from your LAMMPS simulation instead? Screenshots of the structures might be particularly useful. $\endgroup$ May 13, 2023 at 8:48
  • $\begingroup$ How did you pick your forcefield? Have you found any paper or reference where it was applied to Silicon clusters or similar systems? $\endgroup$
    – Greg
    May 14, 2023 at 4:20

1 Answer 1


Most molecular dynamics force fields are set up to model either bulk systems or large unreactive molecules (like proteins). They are not designed for small systems (such as a 13-atom nanoparticle) which are accessible through quantum chemistry methods, and where each atom is likely to have a very different chemical environment which would be very hard to describe using a single functional form for all atoms.


You must log in to answer this question.

Not the answer you're looking for? Browse other questions tagged .