Lammps community,

I am computing the RDF for a periodic 3D fcc lattice using Lennard-Jones. I am seeing a size-dependent effect on my RDF results, and I am wondering why. For a periodic box, the formulation of g(r) should not depend on the number of atoms in the system, yet when I increase nx, ny, and nz in my script (seen below), each g(r) value decreases significantly. In testing this problem, the values eventually converged fully at ~1000000 atoms, but I don’t see why they should change at all with the number of atoms.

# Simulation settings

package gpu 1

boundary p p p

atom_style atomic

units metal# General model information

variable nx equal 1 # number of cells in x direction

variable ny equal 1 # number of cells in y direction

variable nz equal 1 # number of cells in z direction#Create Atoms

lattice fcc 1.0

region box block 0 ${nx} 0 ${ny} 0 ${nz} units lattice

create_box 1 box

create_atoms 1 box

mass 1 1.0# Potential

variable rcut equal 2.5

variable padding equal 2*${rcut}

variable sigAA equal 1.0

variable epsAA equal 1.0

pair_style lj/cut/gpu ${rcut}

pair_coeff 1 1 ${epsAA} ${sigAA}

pair_modify shift yes# Compute radial distribution (RDF) function

comm_modify cutoff ${padding}

compute myRDF all rdf 100 1 1 cutoff ${rcut}

fix fix_rdf all ave/time 1 1 1 c_myRDF[*] file out.rdf mode vector

run 0

I also computed the RDF separately by importing the dump file into OVITO, and did not see a size-dependent effect here – increasing the number of atoms did not have an effect on the g(r) values computed in OVITO (computed with the same number of bins and cutoff).

Any insight into this issue within lammps would be appreciated.

Thank you,

Chloe