Coarse grain model for crystal structure

Hi Lammps users,

I want to use coarse grain model to simulate my clay particle-particle aggregation. But I don’t know how to build this “coarse grain” for my system. I understand how the polymers or membrane are built in the software moltemplate. I am not sure how to define my “monomer”, since all the atoms in my crystal system are shared in 3 dimension. like other crystal structure. I’ve attached my pdb file here. Does anybody know how to use coarse grain model for crystal structure? Many thanks!

kaolinite1224HOH1.pdb (87.7 KB)

Hi Lammps users,

I want to use coarse grain model to simulate my clay particle-particle
aggregation. But I don't know how to build this "coarse grain" for my
system. I understand how the polymers or membrane are built in the software
moltemplate. I am not sure how to define my "monomer", since all the atoms
in my crystal system are shared in 3 dimension. like other crystal
structure. I've attached my pdb file here. Does anybody know how to use
coarse grain model for crystal structure? Many thanks!

before building a coarse grain model, you have to find out a suitable
coarse graining strategy that is applicable to your system at hand. i
doubt that the approach used in polymers can be translated for your
system. and even if it can, you need to find proper parameters for it.
both of which has very little to do with LAMMPS specifically (which is
just an MD code), and the best place to look for answers, if any, is
the published literature.

axel.

Axel,

Thank you very much for your response. I did check CG model applied in layered materials in the literature, although very limited has been published. One way reported is researchers use arrays of silica spheres to represent the silicate sheet, connected by an anharmonic extensible nonlinear elastic potential.(This should be similar to what’s shown in the LAMMPS website–in the non-sphere particles, where the red particles are clumped together to form a rigid body). May I know how this can be accomplished by LAMMPS, or prepared for LAMMPS? Eventually, I want to build two connected sheets, formed by silica sphere and alumina sphere, respectively, I guess this is able to be done after I figure out how to build one layer of silica spheres.

Thanks a lot!

Jing

Axel,

Thank you very much for your response. I did check CG model applied in
layered materials in the literature, although very limited has been
published. One way reported is researchers use arrays of silica spheres to
represent the silicate sheet, connected by an anharmonic extensible
nonlinear elastic potential.(This should be similar to what's shown in the
LAMMPS website--in the non-sphere particles, where the red particles are
clumped together to form a rigid body). May I know how this can be
accomplished by LAMMPS, or prepared for LAMMPS? Eventually, I want to build

i don't know. *you* have read the literature, not me. what is doable
with LAMMPS is listed in the documentation. if the potential functions
offered by LAMMPS match the potential functions used in the
corresponding paper(s), you can use it. otherwise, you'll have to get
out a text editor and start programming (or compel somebody to do the
programming for you), or look for a different MD code. if you have
found a paper describing a simulation, there also has to be a software
capable of doing this kind of model.

two connected sheets, formed by silica sphere and alumina sphere,
respectively, I guess this is able to be done after I figure out how to
build one layer of silica spheres.

mind you, as far as LAMMPS is concerned, it doesn't matter at all what
your model represents. LAMMPS sees just numbers and functions. it
takes the functions, sticks in the numbers and gets a new set of
numbers; rinse and repeat until done. everything else is essentially
is your jobs. LAMMPS only provides the (parallel) platform or
framework, if you want to call it that.

axel.

> Axel,
>
> Thank you very much for your response. I did check CG model applied in
> layered materials in the literature, although very limited has been
> published. One way reported is researchers use arrays of silica spheres
to
> represent the silicate sheet, connected by an anharmonic extensible
> nonlinear elastic potential.(This should be similar to what's shown in
the
> LAMMPS website--in the non-sphere particles, where the red particles are
> clumped together to form a rigid body). May I know how this can be
> accomplished by LAMMPS, or prepared for LAMMPS? Eventually, I want to
build

i don't know. *you* have read the literature, not me. what is doable
with LAMMPS is listed in the documentation. if the potential functions
offered by LAMMPS match the potential functions used in the
corresponding paper(s), you can use it. otherwise, you'll have to get
out a text editor and start programming (or compel somebody to do the
programming for you), or look for a different MD code. if you have
found a paper describing a simulation, there also has to be a software
capable of doing this kind of model.

I know the "potential" part, and LAMMPS is capable to do this. The part I

am not familiar is how to generate those arrays of spheres, maybe I asked
too stupid question, but I really cannot figure this out.

> two connected sheets, formed by silica sphere and alumina sphere,
> respectively, I guess this is able to be done after I figure out how to
> build one layer of silica spheres.

mind you, as far as LAMMPS is concerned, it doesn't matter at all what
your model represents. LAMMPS sees just numbers and functions. it
takes the functions, sticks in the numbers and gets a new set of
numbers; rinse and repeat until done. everything else is essentially
is your jobs. LAMMPS only provides the (parallel) platform or
framework, if you want to call it that.

I am sorry I gave too much information here, my intention is to provide

some information for people trying to help, hoping they may have a better
understanding what I am trying to do. I am afraid I didn't make my question
clear. Thanks anyway.

Jing

axel.

I know the "potential" part, and LAMMPS is capable to do this. The part I am
not familiar is how to generate those arrays of spheres, maybe I asked too
stupid question, but I really cannot figure this out.

fist of all, what you are asking about is very model specific, so you
cannot expect a specific answer that solves your specific problem,
since there likely isn't anybody around that has done what you want to
do, or even knows what model you plan to use.

that said, there has to be a "recipe" how to construct a geometry from
an all atom geometry (so-called mapping) or a strategy to build a CG
geometry from scratch. again, you have to refer to the literature
describing the model as a source of information.

at the end, you'll most likely have to write a script/program that
outputs a data file suitable for LAMMPS to read and the format of that
is explained in the read_data documentation.

[...]

mind you, as far as LAMMPS is concerned, it doesn't matter at all what
your model represents. LAMMPS sees just numbers and functions. it
takes the functions, sticks in the numbers and gets a new set of
numbers; rinse and repeat until done. everything else is essentially
is your jobs. LAMMPS only provides the (parallel) platform or
framework, if you want to call it that.

I am sorry I gave too much information here, my intention is to provide some

not too much, but the wrong information. constructing a geometry is
different from discussing whether a simulation is possible.

i don't know. *you* have read the literature, not me. what is doable
with LAMMPS is listed in the documentation. if the potential functions
offered by LAMMPS match the potential functions used in the
corresponding paper(s), you can use it.

I know the "potential" part, and LAMMPS is capable to do this. The part I am
not familiar is how to generate those arrays of spheres, maybe I asked too
stupid question, but I really cannot figure this out.

I'm really sorry this comes at a bad time. At the moment I am really
busy, and I don't have time to test this before I post it. In the
next few days, I have only a few minutes to write this reply.

I suspect you will have a mixture of sphere (or asphere, or ellipsoid)
particles, and regular point-like particles. I assume you are using
"hybrid full sphere". For more information, see:

I've never tried using atom_style hybrid. Let me know if it
moltemplate gets confused with hybrid atom styles. (Hopefully not!)
I will probably embarrass myself. Here is my attempt to write a
quick and sloppy attempt to explain how to do something like this:

(I hope I don't mess up the column ordering):

--------- sphere.lt ----------

SphereThingy {
  # This "SphereThingy" object contains only one atom.

  # ID type x y z molid charge diameter density
  write("Data Atoms") {
    $atom:x @atom:X 0 0 0 $mol 0 5.0 1.0
  }

  # Comments:
  # particle has zero charge and is located at 0 0 0
  # please change the diameter and density numbers
  # I did not include a "Data Mass" section.
  # Perhaps I should have. Check this.

  write_once("In Settings") {
    pair_coeff @atom:X @atom:X ??? ???
  }
}

# More comments:
# At some point, you must eventually specify the atom_style
# and pair_style. Why not do that here?

write_once("In Init") {
  units real
  atom_style hybrid full sphere
  pair_style ???
}

# You should pick a pair style appropriate for the atom_style

---------- system.lt ------------

import "sphere.lt"

sheet = new SphereThingy
               [10].move(10, 0, 0)
               [10].move(5, 8.660254, 0)

# I don't know how big your sphere's are.
# In this example, I created a 10x10 array of spheres
# hexagonally spaced, with a 10.0 Angstrom spacing.
# Please scale these numbers accordingly.

write_once("Data Boundary") {
  0 100.0 xlo xhi
  0 86.69254 ylo yhi
  0.0 80.00 zlo zhi
}

------------running moltemplate ----------

Then run moltemplate.sh using:

moltemplate.sh -atomstyle "hybrid full sphere" system.lt

(hopefully nothing too bad happens)

I am sorry I gave too much information here, my intention is to provide some
information for people trying to help, hoping they may have a better
understanding what I am trying to do. I am afraid I didn't make my question
clear. Thanks anyway.

I don't have a much good understanding. I will say that it should be
possible to use moltemplate to build systems with unusual particle
types such as spheres

I don't know if the example above makes sense at all, but it's the
best I can do today. Keep me updated if you make any progress or have
any questions.

Good luck

Andrew
P.S.
Be sure to compile LAMMPS with whatever additional packages you need
to support the atom-style you want. I don't know if "atom-style
sphere" is included by default.

(Bug: If you use the "rot()" command, it won't work on ellipsoids or
other asymetric particles. Hopefully that's not a bid deal for most
people. It does not sound like you need to rotate these sheets,
however.)

Andrew,

Thank you very much for your help. I really appreciate it. I will try with this code.

Jing

I forgot to add: For simple periodic sheets like this, you can
probably do this with lammps input scripts commands as well.

Good luck.

Andrew