[lammps-users] Ensemble Regarding

Hi Steven,

Just wanted to know one thing from you. I am trying to simulate nanometric machining phenomena:

Case1: Tool is rigid : Ensemble chosen is micro canonical( nve) which was ok .

Case2: Tool is also subjected to wear: Which ensemble is appropriate to be choosen.

I tried with combinations like all nve, & with tool as nvt & work as nve but am not getting the results.

Can you help me understand the physics behind selection of the kind of ensemble in this particular case.

Regards,

Saurav Goel

I have posted in the earlier thread under topic ensemble, but not sure if it is received by you.

Hi Steven,

Just wanted to know one thing from you. I am trying to simulate nanometric
machining phenomena:

Case1: Tool is rigid : Ensemble chosen is micro canonical( nve) which was ok
.

Case2: Tool is also subjected to wear: Which ensemble is appropriate to be
choosen.

I tried with combinations like all nve, & with tool as nvt & work as nve but
am not getting the results.

Can you help me understand the physics behind selection of the kind of
ensemble in this particular case.

your concerns about "ensemble" are very likely
to miss the point.

you have to look at things from a different perspective.

NVE doesn't mean _enforcing_ a simulation to be in NVE ensemble
as such, but rather it means, all that fix nve does is the plain time
integration; nothing else.

NVT instead does the time integration _and_ manipulates the
acceleration in a way that a bulk system with a finite number of
particles behaves as if it was (homogeneously, i.e. no temperature
gradient) connected to a (near) infinite heat bath. effectively, the
larger the system, the smaller the difference between the two.

however, for your kind of simulation (parts of the system being
manipulated as macroscopic objects) you almost always want
to only do plain time integration during the manipulation (it may
make sense to do individual equilibration with nvt or nve plus
langevin beforehand). if you have a substrate, then you may
want to place a fix langevin on the lower part of the substrate
and above the part that you keep fixed in space to have it
emulate exchange of kinetic energy with the bulk of the substrate
in order to dissipate the energy that is generated on impact
of an object on the substrate.

running that "manipulation" part of the simulation with fix nvt
is almost always a mistake, since you are likely to become
a victim of the "flying icecube syndrom".

in any case, the kind of scenario that you describe is not
really covered by any statistical mechanical ensemble.

cheers,
    axel.

Dear Sir,

Thanks very much for providing the valuable insights. what I understood is that nve ensemble is suited for the case of machining in such a scenario. However, practically in lammps when i am simulating with fix nve, the tool start wearing as that like a liquid & with fix nvt it remain rigid to the larger extent.

What may be the reason.

Regards,
Saurav

Dear Sir,

Thanks very much for providing the valuable insights. what I understood is
that nve ensemble is suited for the case of machining in such a scenario.
However, practically in lammps when i am simulating with fix nve, the tool
start wearing as that like a liquid & with fix nvt it remain rigid to the
larger extent.

What may be the reason.

that is exactly what i would expect and what i was referring to.

you are probably moving your tool at an extremely high
velocity (try to translate it to meters per second then then
compare to the speed of sound, for example).
this adds kinetic energy to the system, so if you run with
nvt that energy is now being removed by the integrator
turning you tool into the "flying icecube" that i mentioned.

think about this for a bit...

cheers,
    axel.

Dear Sir,

Having understood your point, I realized I was using a velocity of 540m/s which is precisely 1.6 times the velocity of sound. Whilst this cutting speed is established normal by the researchers for working MD simulations, i thought ay be it is large enough & Considering your advise, i reduced this to 20m/s with all being considered nve.

However, the result remains same. Further to my surprise, i am unable to understand that why diamond(Carbon) atoms start almost bubbling at start of simulation & are more unstable than silicon.

Regards,
Saurav goel

Dear Sir,

Having understood your point, I realized I was using a velocity of 540m/s
which is precisely 1.6 times the velocity of sound. Whilst this cutting
speed is established normal by the researchers for working MD simulations, i
thought ay be it is large enough & Considering your advise, i reduced this
to 20m/s with all being considered nve.

However, the result remains same. Further to my surprise, i am unable to
understand that why diamond(Carbon) atoms start almost bubbling at start of
simulation & are more unstable than silicon.

do you equilibrate your system, before starting the "experiment"?

please keep in mind that my eyesight does not extend to the
screen of your computer, so i don't exactly know what you are
doing or seeing. hence it is very difficult to discuss about it in
this very abstract way. perhaps the problem is elsewhere.

axel.

Hi,

Trying to simulate nanometric machining of copper against diamond.(attached). the thermo output seems ok but the vmd output loooks horrible. Unable to understand what’s wrong with it. M also surprise to see that there is difference b/w Fx & Fxx. Please advise.

Saurav

copper.trial (3.62 KB)

Sir,

Attached is the modified & seemingly correct program along with log file. I am using Tersoff function & wish to calculate the Fx, Fy & Fz acting on the tool. From the log file, it seems there is something wrong( -ve Fx,Fy Fz).

Reviewing the manual suggest that group/group command for 3 body interaction cannot be used.

What is the way out to calculate the force output / stress output which is appropriate for such a problem.

Regards,
Saurav Goel

tool1.trial (3.34 KB)

log.lammps (9.04 KB)

Sir,

Attached is the modified & seemingly correct program along with log file. I am using Tersoff function & wish to calculate the Fx, Fy & Fz acting on the tool. From the log file, it seems there is something wrong( -ve Fx,Fy Fz).

Reviewing the manual suggest that group/group command for 3 body interaction cannot be used.

What is the way out to calculate the force output / stress output which is appropriate for such a problem.

Regards,
Saurav Goel

tool1.trial (3.34 KB)

log.lammps (9.04 KB)

Attached is the modified & seemingly correct program along with log file. I am using Tersoff function & wish to calculate the Fx, Fy & Fz >acting on the tool. From the log file, it seems there is something wrong( -ve Fx,Fy Fz).
Reviewing the manual suggest that group/group command for 3 body interaction cannot be used.
What is the way out to calculate the force output / stress output which is appropriate for such a problem.

If you have a group of atoms that is the "tool" and then all other atoms,
you can just sum the per-atom force on the tool atoms. That wilil
be the force due to all other atoms. That will work for any
kids of interactions, 3-body, etc.

Steve

Sir,

Attached is the modified & seemingly correct program along with log file. I

when a simulation executes without crashing, that doesn't automatically
mean that it is correct. all it means is that you are not overloading the
implemented algorithms.

now, what is a "correct" model for what you want to find out,
that is a _totally_ different question and one that is much
harder to answer, than where is a problem with overloading
the algorithm.

from looking at the visualization, you have changed your model
from trying to model a solid tool to something that looks more
like a "nano-comb". that is a _very_ different system and you'll
be getting different answers.

but what is the correct model for the property that you want to
compute, is not the concern of the LAMMPS program, but your
own problem.

axel.

Hi,

Thanks for the clarification indeed. Just have one more question. Manual says that the units for the measurement of Compute stress/atom command
will be Pressure*volume Units.

For style metal, these are the units:
• distance = Angstroms
• force = eV/Angstrom
• pressure = bars

Stress is Measured as Force/Area means eV/Angstrom3 on the other hand the conversion would be like

1 eV/Angstrom3 = 160.2176487 GPa while the manual says the pressure is in bar.

whereas 1 bar = 0.000101325 GPa .

This is little confusable. Can you precisely confirm that what is the output measurement unit of Compute Stress/Atom for metal Units??

Regards,
Saurav Goel

Hi,

Thanks for the clarification indeed. Just have one more question. Manual says that the units for the measurement of Compute stress/atom command
will be Pressure*volume Units.

For style metal, these are the units:
• distance = Angstroms
• force = eV/Angstrom
• pressure = bars

Stress is Measured as Force/Area means eV/Angstrom3 on the other hand the conversion would be like

1 eV/Angstrom3 = 160.2176487 GPa while the manual says the pressure is in bar.

whereas 1 bar = 0.000101325 GPa .

This is little confusable. Can you precisely confirm that what is the output measurement unit of Compute Stress/Atom for metal Units??

Regards,
Saurav Goel
HWU, UK

It means bar*Angs^3.

Steve

Hi Steve,

If so is the case then the resultant unit is Newton*Angstorm( which shall be the unit of torque) how come it be converted to stress units?

Regards,
Saurav

I think it would be best for you to work out the details of the units
for yourself. I believe the manual is pretty clear about what the units
of input and output are. If you don't like those units you can
convert them to something else yourself as a pre- or post-processing
step. I suggest you set up 2 atoms in a box with metal units at a known
separation for a simple potential, then print out their pressure,
per-atom stress,
etc, and see if you can verify the numbers LAMMPS spits out, that
they match the units LAMMPS says they are. If you think LAMMPS
is making a mistake (unlikely), then post to the list.

Steve