neibor_modify

LAMMPS LAMMPS Mailing List Mirror
1

Dear Lammps Users & Developers
I am working to simulate shear stress at interface of two different material.
I need to use the “neibor_modify once yes” just at interface but not for all.
I want to know is it possible in LAMMPS to only build neighbor list once at start of run between two specified groups and never rebuild, and also rebuild neighbor list for other atoms?
I studied the documents but I don’t find the solution.
Thanks in advance
Mohammad

2

Dear Lammps Users & Developers
I am working to simulate shear stress at interface of two different
material.
I need to use the "neibor_modify once yes" just at interface but not for
all.
I want to know is it possible in LAMMPS to only build neighbor list once at
start of run between two specified groups and never rebuild, and also
rebuild neighbor list for other atoms?

no. this makes no sense.
if you update the neighbors for one atom, you have to also update the
list of neighbors for all its neighbors or else you will create
massive inconsistencies.

neighbor_modify once only makes sense for solid systems with no
deformation, anyway.

axel.

3

Thanks Axel, maybe I need other method to solve the problem.
when I want to simulate the shear stress at interface, the rupture does not occur.
and I think it is due to during the tangential displacement, the atoms create new bond with other atoms (atoms in the forward). I have read in literature that they have bield the neighbor list just at start of simulation to avoid this problem. I want to know is other technical method to solve this problem in LAMMPS.
Thanks
Mohammad

4

Thanks Axel, maybe I need other method to solve the problem.
when I want to simulate the shear stress at interface, the rupture does not
occur.
and I think it is due to during the tangential displacement, the atoms
create new bond with other atoms (atoms in the forward). I have read in
literature that they have bield the neighbor list just at start of
simulation to avoid this problem. I want to know is other technical method
to solve this problem in LAMMPS.

my suspicion is, that you are misunderstanding the description in the
literature.

also, your approach seems backwards (to put it politely). if atoms get
close, they interact. there is no way in nature to tell some atoms to
suddenly ignore the fact that there are atoms close to it. after all,
you are not trying to do a cartoon here, but want to learn about the
properties of your system, right?

if your model does not behave the way you expect, then either your
model is not set up suitably or your expectations are not correct.
rather than manipulating the model in unphysical ways, you should
first try to understand where exactly the problem is.

axel.

5

Dear Axel

I paste the paragraph of the literature hear. as this project
done with LAMMPS, please give me your suggestion about it.
Thanks
“For mode II crack simulation where a shear load parallel to the crack plane is
applied, the initial crack may be healed when region 1 is
shifted to above region 4 or region 2 is shifted to below
region 3. To overcome this problem, we initiate atom
neighbors at the start of the simulations and do not rede-
termine neighbors. This means that atoms interact only
with the neighbors that are identified at the start of the
loading. It mimics the realistic scenario that once atoms
near the crack tips break their bonds with their neighbors
in the initial crystal configuration, they are quickly contam-
inated (e.g. via oxidation) so that they lose the ability to
reform bonds with new neighbors.”

6

Dear Axel
I paste the paragraph of the literature hear. as this project
done with LAMMPS, please give me your suggestion about it.

it is difficult to judge without the full context. standing on its
own, i personally find the justification questionable, especially for
dynamical processes at atomistic resolution. e.g. the time scale of
oxidation is very different from the time scale of typical MD
dynamics. this is easily rationalized by simply considering the
probability of an oxygen atom hitting a freshly cleaved surface atom.
this kind of explanation would seem more reasonable for a continuum
model with different time and size scales.

on top of that, the change in atom properties would be much better
modeled by changing the atom type to a potential that is less
attractive than through manipulations of the neighbor list.

if you look at the "crack" example in the LAMMPS distribution, you can
see how a neighborlist manipulation is used properly. its purpose is
solely to facilitate a crack formation at a specific location. the
"science" happens only in the part where there are no such
manipulations.

axel.

7

Thanks for your reply.
I know the method of “crack” example and this literature used this method to create a local crack.

but the question is exist yet.

for instant if the “shear” example run at P P P boundary condition and let to run more time

the rupture don’t occur and the crystal rebuild itself due to create new bond.

I want to know how I can simulate the shear and rupture in periodic direction.

Thanks in Advance

Mohammad

8

Thanks for your reply.
I know the method of "crack" example and this literature used this method to
create a local crack.
but the question is exist yet.
for instant if the "shear" example run at P P P boundary condition and let
to run more time
the rupture don't occur and the crystal rebuild itself due to create new
bond.
I want to know how I can simulate the shear and rupture in periodic
direction.

using P P P on the shear example makes no sense at all.

using P S P means you change the model and for material that you push
out of the box, you get new material that enters the box. if you shear
your system along a periodic axis, then it would be the correct model
to have the new material coming in interact with what is already
there. if i run the simulation, i see nothing unexpected.
what is the physical reason, why atoms should behave differently?

axel.

9

when I want to simulate the shear stress at interface, the rupture does not occur.
and I think it is due to during the tangential displacement, the atoms create new bond with other atoms (atoms in the forward). I

What you are describing here is plastic deformation through nucleation and movement of (edge) dislocations. I don’t see why you would need to mess with neighbor lists for this type of simulation.

10

when I want to simulate the shear stress at interface, the rupture does not occur.
and I think it is due to during the tangential displacement, the atoms create new bond with other atoms (atoms in the forward). I

What you are describing here is plastic deformation through nucleation and movement of (edge) dislocations. I don’t see why you would need to mess with neighbor lists for this type of simulation.
Daniel