Granular simulations-stress/strain

Hi all,

I am trying to simulate the stress-strain behavior of granular particles in a simulation box. When I try to apply load for a longer duration, the atoms get lost. I would like to know if there is a way to do this without the atoms getting lost and still observe the mechanical behavior(plastic/elastic) of the particles.

Regards,
Muthukumar

There is not much information here. Please include at least a sample of your i/p script, which tells the potential,timestep,strain rate, command used to achieve deformation etc.

Sagar

Hi all,

I am trying to simulate the stress-strain behavior of granular particles in
a simulation box. When I try to apply load for a longer duration, the atoms
get lost. I would like to know if there is a way to do this without the
atoms getting lost and still observe the mechanical
behavior(plastic/elastic) of the particles.

lost atoms are almost invariably a sign of a bad choice of simulation
parameters.

axel.

Please find the sample input script

Initialization

units si
boundary f f p
atom_style sphere
dimension 2
newton off off
communicate single vel yes

Atom Definition

region openbox block 0 0.01 0 0.01 0 14e-6 units box
create_box 1 openbox
create_atoms 1 random 500 3 openbox units box
set region openbox diameter 14e-6 density 1700

#Settings
neighbor 15e-5 bin
#neighbour modify
pair_style gran/hertz/history 146520146520 180995475113 5e-5 0 0.5 0
pair_coeff * *
timestep 0.000000000005
run_style verlet
fix 2 all wall/gran 146520146520 180995475113 5e-5 0 0.5 0 yplane 0 0.01
fix 3 all nve/limit 6e-6
fix 4 all momentum 1 angular
fix force all indent 3 plane y 0.05 lo units box
compute new all temp
compute 1 all pressure new
dump 1 all xyz 5 output1.xyz
thermo 100
thermo_style custom step press
run 65000

Please find the sample input script

# Initialization
units si
boundary f f p
atom_style sphere
dimension 2
newton off off
communicate single vel yes

# Atom Definition
region openbox block 0 0.01 0 0.01 0 14e-6 units box
create_box 1 openbox
create_atoms 1 random 500 3 openbox units box
set region openbox diameter 14e-6 density 1700

#Settings
neighbor 15e-5 bin
#neighbour modify
pair_style gran/hertz/history 146520146520 180995475113 5e-5 0 0.5 0
pair_coeff * *
timestep 0.000000000005

Too high a timestep..!!!!! Try making it 5e-15.

You might perform a Buckingham-Pi analysis and choose simulation parameters accordingly. There is a wealth of studies performed in this area that have already worked out the limits of convergence over the past 20-30 years, and its much easier to inspect parameters and compare simulations when properly non-dimensionalized. Likewise sensitivity studies (of which there are many) will also show you that you need not necessarily concern yourself with matching all the physical parameters you want to study. For example, simulating realistically hard particles results in a separation of scales. You have to resolve an overlap which is orders of magnitude smaller than the scale of the particle, which results in very slow simulations, at the same time stresses do not change appreciably by lowering your spring constant.