Hi, I have run NEMD thermal conductivity calculation for silicon nanowires using M-P method. I was able to complete the simulation with reasonable thermal conductivity data. However, as I visualized the nanowire structure with OVITO, I found that the wire was rotating along the propagation direction during the NVE equilibration and thermal conductivity calculations. I wonder whether or not it’s normal since the dimension of the wire itself doesn’t change while rotation. I am attaching a part of my input file and the images of the wire (looking from the propagation direction) rotating during the simulations.
Thanks in advance,
Hi, I have run NEMD thermal conductivity calculation for silicon nanowires
using M-P method. I was able to complete the simulation with reasonable
thermal conductivity data. However, as I visualized the nanowire structure
with OVITO, I found that the wire was rotating along the propagation
direction during the NVE equilibration and thermal conductivity
calculations. I wonder whether or not it's normal since the dimension of the
wire itself doesn't change while rotation. I am attaching a part of my input
file and the images of the wire (looking from the propagation direction)
rotating during the simulations.
if there are no interactions with periodic images, it is quite likely
that a rod/wire shape object will pick up angular momentum due to
forces not perfectly balancing as a result of floating point
rounding/truncation issues, this would be particularly the case, if
the object is not centered on the origin.. also, an initial momentum
can be imprinted on the system, when initial velocities are applied
and residual momenta not removed.
Thanks for the replies. My structure seems to be centered inside the box. As I can see that the nanowire starts rotating before M-P calculation starts, i think there is something else. Since I set ‘mom’ and ‘rot’ options to ‘yes’ in ‘velocity’ command, I wonder if any initial velocities and momenta are left in the system. I am copying more lines below from my input file, so hopefully I can hear more suggestions from Lammps users.
Thank you in advance.
Thanks for the replies. My structure seems to be centered inside the box. As
whether the object is centered in the box is irrelevant, but whether
the object's center coincides with the origin (and whether the object
has a symmetry that is commensurate with a cartesian coordinate
system) matters. if the latter two are not given you can pick up
residual forces that may result in either translation (most common) or
rotation. this depends a bit on system size and also the nature and
implementation of the potential.
if you want to know some more where this is coming from, you can
google for the text "what every computer scientist should know about
floating point arithmetic".
I can see that the nanowire starts rotating before M-P calculation starts, i
think there is something else. Since I set 'mom' and 'rot' options to 'yes'
in 'velocity' command, I wonder if any initial velocities and momenta are
left in the system. I am copying more lines below from my input file, so
hopefully I can hear more suggestions from Lammps users.
partial inputs are rarely helpful. murphy's law stipulates that for
most cases the problem in the parts that are not copied.
in general, you need to understand where the momentum is coming from.
i've given a few possible reasons, those can be counteracted to
eliminate or reduce the undesired rotation. if all else fails, you may
look into fix momentum.
