High Spatial Correlation

Hello LAMMPS users/developers!

I’m tryring to simulate a 10000 LJ-Ar atoms system considering that each particle is a neighbour of the other 9999.

This aims to represent a system with maximum spatial correlation (as near around the Critical Point).

I’ve chosen cut-off radius as half of the simmulation box length.

I had no problem to simulate a 1,000 particles system. But i cannot simulate a 10,000 one. And i intend to simulate a 100,000 particles system.

I’ve augmented “page” and “one” as suggested in neigh_modify command in LAMMPS’ Documentation and Mail List to great values like 100,000 or 1,000,000.

But i only have this error message: Neighbor page size must be >= 10x the one atom setting (neighbor.cpp:199)

Do you have any hint to help me in proceed my simmulations?

Best regards!

Jakler Nichele
Military Institute of Engineering

Defense Engineering Graduate Program
Rio de Janeiro - Brazil

Hello LAMMPS users/developers!

I'm tryring to simulate a 10000 LJ-Ar atoms system considering that each
particle is a neighbour of the other 9999.

This aims to represent a system with maximum spatial correlation (as near
around the Critical Point).

I've chosen cut-off radius as half of the simmulation box length.

I had no problem to simulate a 1,000 particles system. But i cannot simulate
a 10,000 one. And i intend to simulate a 100,000 particles system.

I've augmented "page" and "one" as suggested in neigh_modify command in
LAMMPS' Documentation and Mail List to great values like 100,000 or
1,000,000.

But i only have this error message: Neighbor page size must be >= 10x the
one atom setting (neighbor.cpp:199)

Do you have any hint to help me in proceed my simmulations?

the error message is pretty clear about where your
input appears to be failing. if you think that lammps is
at error here and your input correct, please provide an
input to demonstrate it.

that all being said, it is not quite obvious to me, how
all this massive effort will result in something useful.
have you looked at how the ewald/n package can handle
long(er)-range interactions including lennard jones r^6
terms more efficiently?

cheers,
    axel.

Dear Axel,

I definitely don’t think that LAMMPS has an error at this point. I know that my input is failing somewhere.

Then I send it to you. Could you please take a look?

Thanks

Jakler Nichele
Military Institute of Engineering

Defense Engineering Graduate Program
Rio de Janeiro - Brazil

Ar-NPT-10k-160-50.txt (1.14 KB)

Dear Axel,

I definitely don't think that LAMMPS has an error at this point. I know that
my input is failing somewhere.

Then I send it to you. Could you please take a look?

there is nothing that stands out.

however, if you actually want to loop over all pairs
of atoms, it may not be that good an idea to choose
a code like lammps that uses verlet lists. you won't
really benefit from them, but require a lot of memory
to create them.

it may be better to stick with a very simple minded
or even self-written MD code that doesn't use neighbor
lists at all.

axel.

in addition, it is probably worth mentioning,
that a cutoff of larger than 100 angstrom is pretty
much meaningless for a 12-6 LJ potential.
double precision floating point numbers have
only about 14 significant digits accuracy.

axel.