tabulated potential with large gradient in short range distance

LAMMPS LAMMPS Mailing List Mirror
1

Dear all,

I tried to simulate alpha quartz with a special potential proposed in this paper http://www.sciencedirect.com/science/article/pii/S0022309315301253.

I linearly tabulated potential in two ways: Firstly, I use log scale of distance (from 10^i for i=-3 to 1 by 0.0004, giving 10001 data points); Also, I considered even space in the square of distance (r^2 from 0.000001 to 99.800001 by 0.01, giving 9981 data points). I used pair_write to see how well the interpolation but for distance less than 0.1, interpolations from both setting are very bad.

The potential table is created by Mathematica with a log scale distance.

Can I ask how to improve the interpolation in short range with large gradient?

Many thanks

Bingyu Cui

#Initialization############################################################

units metal

dimension 3

boundary p p p

atom_style charge

Atom definition##########################################################

lattice custom 5.4047 &
a1 0.9092 0.0000 0.0000 &
a2 -0.4546 0.7873 0.0000 &
a3 0.0000 0.0000 1.0000 &
basis 0.4697 0.0000 0.0000 &
basis 0.0000 0.4697 0.6667 &
basis 0.5303 0.5303 0.3333 &
basis 0.4133 0.2672 0.1188 &
basis 0.2672 0.4133 0.5479 &
basis 0.7328 0.1461 0.7855 &
basis 0.5867 0.8539 0.2145 &
basis 0.8539 0.5867 0.4521 &
basis 0.1461 0.7328 0.8812

region simbox block 0 2 0 2 0 2 units lattice

create_box 2 simbox

create_atoms 1 box basis 1 1 basis 2 1 basis 3 1 basis 4 2 basis 5 2 basis 6 2 basis 7 2 basis 8 2 basis 9 2

mass 1 28.0855

mass 2 15.9994

group siliconatoms type 1

group oxygenatoms type 2

set group siliconatoms charge 2.4

set group oxygenatoms charge -1.2

Atoms interactions settings##################################

Si type 1, O type 2

pair_style table linear 10001

pair_coeff 1 1 potential.table SiSi

pair_coeff 1 2 potential.table SiO

pair_coeff 2 2 potential.table OO

neigh_modify every 1 delay 0 check yes

pair_write 1 2 10001 r 0.001 10.0 table.txt Si_O 2.4 -1.2

pair_write 2 2 10001 r 0.001 10.0 table.txt O_O -1.2 -1.2

pair_write 1 1 10001 r 0.001 10.0 table.txt Si_Si 2.4 2.4

Equilibrate

fix 1 all npt temp 1986.15 5200.0 0.1 iso 1.0 1.0 1

thermo_style custom step temp pe etotal press pxx pyy pzz vol
thermo 10

dump 1 all atom 1000 dump1.qua
timestep 0.001

run 1000000

unfix 1

fix 1 all npt temp 5200.0 5200.0 0.1 iso 1.0 1.0 1

thermo_style custom step temp pe etotal press pxx pyy pzz vol
thermo 10

dump 2 all atom 1000 dump2.qua
timestep 0.001

run 10000

unfix 1

fix 1 all box/relax iso 0.0 vmax 0.001

thermo_style custom step temp pe etotal press pxx pyy pzz vol
thermo 10

dump 3 all atom 1000 dump3.qua
timestep 0.001

run 1000000

min_style cg

minimize 1e-25 1e-25 5000 10000

2

Bingyu Cui has shared OneDrive files with you. To view them, click the
links or images below.
<https://1drv.ms/u/s!Atz7WaJYYMWir2ct0a-IgyrS90hs>
<https://1drv.ms/u/s!Atz7WaJYYMWir2ct0a-IgyrS90hs>

<https://1drv.ms/t/s!Atz7WaJYYMWir2QA4ACApgBAg-Ro>
table.txt <https://1drv.ms/t/s!Atz7WaJYYMWir2QA4ACApgBAg-Ro>
<https://1drv.ms/t/s!Atz7WaJYYMWir2QA4ACApgBAg-Ro>

<https://1drv.ms/u/s!Atz7WaJYYMWir2Uems05ECq2lkgo>
in.alphaquartz <https://1drv.ms/u/s!Atz7WaJYYMWir2Uems05ECq2lkgo>
<https://1drv.ms/u/s!Atz7WaJYYMWir2Uems05ECq2lkgo>

<https://1drv.ms/u/s!Atz7WaJYYMWir2ZeJxm4X1x9bFyp>
potential.table <https://1drv.ms/u/s!Atz7WaJYYMWir2ZeJxm4X1x9bFyp>
<https://1drv.ms/u/s!Atz7WaJYYMWir2ZeJxm4X1x9bFyp>

Dear all,

I tried to simulate alpha quartz with a special potential proposed in this
paper http://www.sciencedirect.com/science/article/pii/S0022309315301253.

I linearly tabulated potential in two ways: Firstly, I use log scale of
distance (from 10^i for i=-3 to 1 by 0.0004, giving 10001 data points);
Also, I considered even space in the square of distance (r^2 from 0.000001
to 99.800001 by 0.01, giving 9981 data points). I used pair_write to see
how well the interpolation but for distance less than 0.1, interpolations
from both setting are very bad.

​please look up the diameter for the atoms in your system and typical or
minimal bond lengths for the conditions of your system. i'd wager a guess,
that those atoms would never come as closes as 1.0 angstrom, let alone 0.1
\AA. so, if the tabulation is only bad for such short distances, then i
would not worry.

axel.

3

Dear Dr. Axel,

Thank you for your quick answer. Yes, the actual atomic sizes of silicon and oxygen should prevent them from getting as close as 0.1 \AA However, when I instead interpolate tabulate potential from 10^-1 \AA, then lammps reported an error that pair distance is less than table inner cutoff. Further, when writing input script, I only set atomic mass and charge without bond nor atomic size.

How can we know short range potential (< 1\AA) will not be used?

Thanks

Bingyu Cui

4

Dear Dr. Axel,

Thank you for your quick answer. Yes, the actual atomic sizes of silicon
and oxygen should prevent them from getting as close as 0.1 \AA However,
when I instead interpolate tabulate potential from 10^-1 \AA, then lammps
reported an error that pair distance is less than table inner cutoff.
Further, when writing input script, I only set atomic mass and charge
without bond nor atomic size.

How can we know short range potential (< 1\AA) will not be used?

​from knowing the physics of your system.

axel.​