How to establish a model with both granular particles and atoms in LAMMPS

Dear lammps users,

I wonder whether it is possible to establish a model with both granular aggregates and atoms in LAMMPS. It seems very tricky to define atom_style and bond_style for particles and atoms, respectively. Besides, the bond (or contact) between particle and atoms seems very difficult to define, and I didn’t find any relevant command for the particle-atom contact.

Does anybody have a good suggestion on solving these issues?

Thanks in advance for your kind help!

Wenjuan

Dear lammps users,

I wonder whether it is possible to establish a model with both granular
aggregates and atoms in LAMMPS. It seems very tricky to define atom_style
and bond_style for particles and atoms, respectively. Besides, the bond (or
contact) between particle and atoms seems very difficult to define, and I
didn't find any relevant command for the particle-atom contact.

Does anybody have a good suggestion on solving these issues?

before entering this rabbit hole, you first need to explain what it is
you want to model with this approach. the granular models in LAMMPS
are meant to represent particles that are several order of magnitude
larger in size than atoms. from the perspective of an atom, the
properties of those granular particles make no sense. you could just
as well simulate a slab.

axel.

If you want dynamics (of the granule) you might as well solve a langevin equation. If you’re interested in surface effects (e.g. double layer), you might find this demands quite a few atoms, as per Axel’s comment.

Dear Axel and Eric,

Thanks for the quick response. What I try to do is to establish a model in LAMMPS to bridge atomic scale and microscale/mesoscale. For example, the crack initiation and propagation in a bulk may be originated from atomic bond break and then gradually nucleate into a crack at microscale/mesoscale that human eye can see. The interface region between the atoms atoms and particles may be bridge by some virtual atoms as coarse grained model or some other possible techniques. I am more interested in the simulation of modeling the crack failure of a bulk from both atomic scale and microscale.

Besides, the MD simulation and DEM simulation are pretty similar to each other due to the discrete nature of atomic structure and inter-particle contacted structure, and the movements of both atoms and particles follow Newton’s second law. Due to the inherent similarity between atomic structure and discrete structure consisting of particles, I thought it might be possible to relate MD model and DEM model. That is to say that I wonder whether it is possible to define both inter-atom potential and inter-particle contact, as well as inter-atom-particle contact in one simulation in LAMMPS. Would you mind to give me any suggestions?

Thank you so much!

Wenjuan

How would you couple DEM with MD, if you’re concerned about granules at the atomistic level? DEM employs models with several levels of coarse graining, e.g. atomistic to continuum and then continuum to DEM through resultant force laws from continuum mechanics. At each level you lose more information; there is no atomistic info at the DEM level, you can only get kinetics from these interactions. If you’re concerned with how a crack nucleates, I assume you want to model more than surface forces. Exploit your scale separation and do some modelling.

Dear Axel and Eric,

Thanks for the quick response. What I try to do is to establish a model in
LAMMPS to bridge atomic scale and microscale/mesoscale. For example, the
crack initiation and propagation in a bulk may be originated from atomic
bond break and then gradually nucleate into a crack at microscale/mesoscale
that human eye can see. The interface region between the atoms atoms and
particles may be bridge by some virtual atoms as coarse grained model or
some other possible techniques. I am more interested in the simulation of
modeling the crack failure of a bulk from both atomic scale and microscale.

this is a rather pointless effort. you don't seem to have your time
and length scales figured out.

if we say a granular particle would move at a speed of 1 m/s (which is
quite fast, right?),
this would translate into 1nm/ns or 0.001 nm/ps or 0.000001 nm/fs or
0.00001 angstrom/fs
which is the time scale of your atomic simulation. 100,000 time steps
to have something move by one angstrom on the time and length scale of
atomic particles is an eternity.

secondly, the size aspect. a typical granular particle is so large
from the point of view of a single atom, that it would appear to be
infinitely large (and heavy), so there is not much of a point in
moving it or considering its shape or anything related.

Besides, the MD simulation and DEM simulation are pretty similar to each
other due to the discrete nature of atomic structure and inter-particle
contacted structure, and the movements of both atoms and particles follow
Newton's second law. Due to the inherent similarity between atomic structure
and discrete structure consisting of particles, I thought it might be
possible to relate MD model and DEM model. That is to say that I wonder
whether it is possible to define both inter-atom potential and
inter-particle contact, as well as inter-atom-particle contact in one
simulation in LAMMPS. Would you mind to give me any suggestions?

the fact that you can solve the same equations of motion for both
kinds of particles doesn't mean that you can just make a direct
connections between the two. the challenge that you are facing is
similar to what for example the folks programming the
Atom-to-Continuum package dealing with. and they usually don't need to
consider such a large difference in scales.

i think that before even starting to consider, how this multi-scale
modeling can be implemented for real in a code like LAMMPS, you need
to spend some serious time pondering the physics of the coupling you
are trying to do, and (hopefully) you will see that it is a far larger
challenge than what you make it to be now.

axel.

Dear Axel and Eric,

Thank you so much for the detailed explanation.

If it is not meaningful to conduct simulation of DEM model and MD model in one simulation. How about analysis of microscale failure using particle and then further analysis of interested region using atoms at atomic scale? To concurrently communicate information between these two models at two scales in LAMMPS, is it possible to implement some API or some commands, or some programs in LAMMPS?

Thanks a lot!

Wenjuan

Dear Axel and Eric,

Thank you so much for the detailed explanation.

If it is not meaningful to conduct simulation of DEM model and MD model in
one simulation. How about analysis of microscale failure using particle and
then further analysis of interested region using atoms at atomic scale? To

to learn what?

concurrently communicate information between these two models at two scales
in LAMMPS, is it possible to implement some API or some commands, or some
programs in LAMMPS?

this is not a question of the API, but rather a question of deriving a
model that can describe this kind of coupling. you can add as many
APIs as you want, if you don't have an model to implement and an
algorithm to do it, it is a futile effort.

please stop asking about technical details before you have your
physics sorted out. you are doing the process backwards. i already
mentioned the Atoms-to-Continuum package as an example of coupling to
different kinds of models. have a look at the papers describing its
foundations, derivation and the implementation and you will
(hopefully) see what i am talking about.

thanks,
    axel.

Dear Axel,

Thanks for the suggestion! Appreciated!

Wenjuan