Buckingham and Tersoff to model Silica

Lammps users,

I am modeling alpha quartz with Tersoff and buckingham (famous BKS paper). I am finding when I run the same data file, I get very different simulation size boxes. Shouldn’t I expect that when modeling alpha quartz with Tersoff and buckingham should produce a similar simulation box size?

Ben

Have you visualized you system and see if which one is making more sense in terms of structure. When you say very different, how different they are?

When I minimize both of them and just run under NPT, the tersoff simulation has volume of around 18000 in units metal and the buckingham has around 14000 in units metal. If you’d like to see my simulation script, I attached it below. It is very basic currently since I just want to make sure I am using both potentials accuractely:

Tersoff:

units metal
atom_style full

#read_restart data.restart1
read_data SiO2.data

pair_style tersoff
pair_coeff * * …/…/…/…/potentials/SiO.tersoff O Si

thermo_style custom step temp pe ke etotal press vol
thermo_modify norm yes
#Minimize the potential

thermo 1
fix 1 all nve
fix 2 all box/relax iso 0.0 vmax 0.001
min_style cg
timestep 0.001
#Minimize the potential
thermo 1
fix 1 all nve
fix 2 all box/relax iso 1.0 vmax 0.001
minimize 1.0e-12 1.0e-12 1000 10000
unfix 1

fix 21 all npt temp 300.0 300 0.1 iso 1 1 1
run 10000

Buckingham:

units metal
atom_style charge
kspace_style pppm 1.0e-4
kspace_modify gewald 0.29
neighbor 1.0 bin
neigh_modify delay 0 every 1 check yes

#read_restart data.restart1
read_data SiO2.data

pair_style buck/coul/long 10.0 12.0
pair_coeff 1 1 1388.77 .3623188 175.0
pair_coeff 1 2 18003 .2052124 133.5381
pair_coeff 2 2 0 .1 0

thermo_style custom step temp pe ke etotal press vol
thermo_modify norm yes

timestep 0.001
#Minimize the potential
thermo 1
fix 1 all nve
fix 2 all box/relax iso 1.0 vmax 0.001
minimize 1.0e-12 1.0e-12 1000 10000
unfix 1

fix 21 all npt temp 300.0 300 0.1 iso 1 1 1
run 10000

Where you able to reproduced any of the published data? That should be your 101 test every single time you are new to a topic and would like to have some peace of mind related to knowing what you are doing.

Carlos

I know that would indeed be my next step. I was just checking to see if anyone on this forum had an intuitive reason as to why the simulation box size would be so different for different potentials. I know that I could determine which one is working better by comparing with experimental data. I am just not sure why the simulation box size would be so different. Even if I ran calculations to find say, the density. Obviously at least one of them is wrong and so I was not sure why my implementation was not correct.

Ben

I know that would indeed be my next step. I was just checking to see if
anyone on this forum had an intuitive reason as to why the simulation box
size would be so different for different potentials.

Why trust somebody else before trying on your own? Your "potential savior"
could easily be less qualify than you to judge what the problem is.

I know that I could determine which one is working better by comparing
with experimental data.

NO NO NO. What you need to reproduce is the published data for each
potential. Both parametrizations could have been done with different
applications in mind.

I am just not sure why the simulation box size would be so different. Even
if I ran calculations to find say, the density. Obviously at least one of
them is wrong and so I was not sure why my implementation was not correct.

See above. You actually have no proof that any of them is correct until you
reproduce some published data.
Carlos

I know that would indeed be my next step. I was just checking to see if
anyone on this forum had an intuitive reason as to why the simulation box
size would be so different for different potentials. I know that I could

that is the wrong logic.

*first* you have to make sure that *each* of your calculation is
correct; *then* you can worry about differences, if any.

determine which one is working better by comparing with experimental data. I

wrong again. you do *not* compare to experimental data, but to
published data bundled with the potential at hand that describes the
results of *that potential*. they are very likely *different* from
experimental data. hell, most DFT calculations are likely to be off.

am just not sure why the simulation box size would be so different. Even if
I ran calculations to find say, the density. Obviously at least one of them
is wrong and so I was not sure why my implementation was not correct.

if you test each input individually, you don't have to guess. ...and
*never* exclude the fact, that you may have a typo or some other
logical mistake in the input that may turn *both* calculations wrong.

axel.