Issues of Simulating High-velocity Collision
System: 1 high-velocity atom collides with 4 atoms that are relatively stationary at beginning.
Method: MD using NVE and LJ force field.
We did two cases:
-
When the high-velocity atom collides with a velocity of 0.2 A/fs, the four stationary atoms move due to the collision.
-
When the high-velocity atom collides with a velocity of 50 A/fs, the four atoms remain stationary after collision, just like nothing happened.
Question: We suspect this was caused by the inadequate precision of NVE. Is this understanding correct? Can this be fixed?
The attachment includes the input/data files and the PDB files for the two cases.
Test.rar (73.5 KB)
Issues of Simulating High-velocity Collision
System: 1 high-velocity atom collides with 4 atoms that are relatively stationary at beginning.
Method: MD using NVE and LJ force field.
We did two cases:
-
When the high-velocity atom collides with a velocity of 0.2 A/fs, the four stationary atoms move due to the collision.
-
When the high-velocity atom collides with a velocity of 50 A/fs, the four atoms remain stationary after collision, just like nothing happened.
Question: We suspect this was caused by the inadequate precision of NVE. Is this understanding correct? Can this be fixed?
No. My guess would be the reason is an inadequate choice of time step.
The attachment includes the input/data files and the PDB files for the two cases.
Nobody here cares to deal with rar archives.
Hi Axel
I have narrowed the timestep very small, it doesn’t work. And I found you can not access the rar archives. So do you know how to attach the files?
Hi Axel
I have narrowed the timestep very small, it doesn't work. And I found you
can not access the rar archives. So do you know how to attach the files?
i *do* know how to attach files, yes.
i also know how to extract rar files, but unlike gzip tar files, they
require using a special tool. please keep in mind that most of the
LAMMPS developers use linux or macosx machine where tools that are
popular on windows are not common or not available.
in any case, i looked at your inputs and they show that you are asking
for trouble.
you should not use an NVT time integrator on your simulation.
especially not for those projectile particles. that simply doesn't
make sense, their energy is set by the impact velocity and should not
be coupled to any kind of thermostat. doing this would be very
unphysical. furthermore, those extremely high differences in
temperature (and temperature doesn't really have a meaning for a
system like yours in the first place).
in case of the high-velocity example the resulting temperature for a
system like yours would be half a *billion* kelvin.
if you switch to NVE time integration, you will see that there is an
impact, but - for geometric reasons - relative to the velocity of the
projectile, very little energy is transferred.
axel.
Hi Axel
I first used NVT for the equilibrium of the structure and when I apply the projectile atoms velocity, I unfix the NVT. And i think when I give a high velocity of projectile atoms
it will result in a pretty high temperature. But I am confused just what you say that when the projectile atoms impact the atoms, it should transfer the kinematic energy to
the atoms. But the atoms don’t move. It is unphysical. So do you think is it the NVE algorithm problem or other problems?
Hi Axel
I first used NVT for the equilibrium of the structure and when I apply the
doing an NVT equilibration for a handful of isolated atoms is outright
ridiculous in any case. doing it for a few steps even more so.
projectile atoms velocity, I unfix the NVT. And i think when I give a high
velocity of projectile atoms
it will result in a pretty high temperature. But I am confused just what you
say that when the projectile atoms impact the atoms, it should transfer the
kinematic energy to
it does. but remember that these are not hard spheres, but soft
particles. so the amount of energy transferred depends on the amount
of overlap and repulsion between the particles and the residence time.
increase the diameter, i.e. sigma and you will see.
the atoms. But the atoms don't move. It is unphysical. So do you think is it
they do move.
the NVE algorithm problem or other problems?
the problem is in your head. the simulation does exactly what i would
expect for the input and system you have defined.
axel.