Hi,
I am doing some testing with the granular package. I take a bunch of spherical particles in a box.
atom_style sphere
boundary p p p
newton off
comm_modify vel yes
region box block 0 81 0 81 0 81 units lattice
create_box 1 box
create_atoms 1 random 200000 35418 box
set group all diameter 1
set group all mass 1
Then I use gran/'hooke pair style but I keep only spring part of the normal force, others are set to zero.
pair_style gran/hooke 200000 NULL 0.0 NULL 0.0 0
pair_coeff * *
I run it with nve/sphere and expect the energy of the system to be fluctuating around some mean value because the dissipative part of the force is zero. But I observe a systematic increase in the energy of the system.
timestep 0.0001
fix 1 all nve/sphere
thermo 1000
run 10000000
Can anyone give me some explanation for this? Am I making any mistake somewhere?
logfile and plot of the energy are attached.
With thanks and regards,
Vinay Vaibhav
Ph.D. Student, Theoretical Physics
The Institute of Mathematical Sciences Chennai, India
energy.pdf (27.7 KB)
log.lammps (272 KB)
Hi,
I am doing some testing with the granular package. I take a bunch of
spherical particles in a box.
atom_style sphere
boundary p p p
newton off
comm_modify vel yes
region box block 0 81 0 81 0 81 units lattice
create_box 1 box
create_atoms 1 random 200000 35418 box
set group all diameter 1
set group all mass 1
Then I use gran/'hooke pair style but I keep only spring part of the normal
force, others are set to zero.
pair_style gran/hooke 200000 NULL 0.0 NULL 0.0 0
pair_coeff * *
I run it with nve/sphere and expect the energy of the system to be
fluctuating around some mean value because the dissipative part of the force
is zero. But I observe a systematic increase in the energy of the system.
timestep 0.0001
fix 1 all nve/sphere
thermo 1000
run 10000000
Can anyone give me some explanation for this? Am I making any mistake
somewhere?
have you tried a smaller time step?
non-conservation of energy is frequently caused by using too large a time step.
axel.
Likewise, using the create_atoms random protocol isn’t a great way to prepare a granular system at 20% volume fraction. A quick back of the envelope calculation shows that the mean nearest neighbor distance for a Poisson process is /D~0.78 (overlap/D of ~0.22). That is huge for a granular particle! Any overlap greater than 0.01 is generally inappropriate for simulating commonly encountered granular materials.