Was the Neighbor Multi style deprecated? It’s still in the documentation, however attempting to use it in a script results in a style does not exist error.
On a related question, is there a way to force a shorter neighbor length than the diameter of the largest particle, with atom_style sphere? I have a single large particle and many smaller ones, so it’s unnecessary pair particles as far out as 2x the radius of the largest particle.
Was the Neighbor Multi style deprecated? It’s still in the documentation, however attempting to use it in a script results in a style does not exist error.
no. it is working fine for me with the colloid example and LAMMPS version 7 Aug 2019.
please let us know: the LAMMPS version you are using and the exact command line you use to launch LAMMPS.
please also, if possible, provide a minimal input example to reproduce the issue.
On a related question, is there a way to force a shorter neighbor length than the diameter of the largest particle, with atom_style sphere? I have a single large particle and many smaller ones, so it’s unnecessary pair particles as far out as 2x the radius of the largest particle.
please let us know how you come to the conclusion that the particle diameter is considered and provide us with an example to reproduce it.
thanks,
axel.
Thanks for the reply, I’m using 7 Aug 2019 LAMMPS.
The cutoff is calculated, as I understand it, to be whichever is largest, long-range force cutoff, or 2x the radius of the largest particle plus the skin depth? When I run simulations with similar characteristics to the one provided below I get the output:
master list distance cutoff = 5.1
ghost atom cutoff = 5.1
However because there is only one particle with diameter 5, the longest cutoff distance I actually need is 3.25.
I run using ‘$ mpirun -np 8 lmpPerm_mpi -echo both -i LAMMPSNeighborMulti.in’ Here is an example script that is pretty paired down:
atom_style hybrid sphere charge
boundary p p p
newton off
atom_modify sort 1000 10
comm_modify mode multi vel yes
units si
processors 2 2 2
#partition yes 2 processors 2 2 2
lattice sc 2
region reg block -100 100 -100 100 -100 100 units box
create_box 2 reg
neighbor 0.1 multi
neigh_modify delay 0
pair_style hybrid/overlay granular coul/long 2
pair_coeff * * granular dmt 1e6 0.1 .3 0.00286 tangential linear_nohistory 1 0.4 rolling sds 10 0.4 0.8 damping tsuji
pair_coeff * * coul/long
create_atoms 1 single 0 0 0
set type 1 diameter 5 density 3000
create_atoms 2 single 0 0 5
set type 2 diameter 0.5 density 3000
dielectric -1
kspace_style pppm 50
kspace_modify force/disp/real 5e-3 force/disp/kspace 2.5e-3 mesh 100 100 100
variable massCharge atom mass*sqrt(7.425826298e-21)
set group all charge v_massCharge
run_style respa 3 10 15 kspace 3 pair 1
fix int all nve/sphere
timestep .5
run 100
Thanks for your help,
Thanks for the reply, I’m using 7 Aug 2019 LAMMPS.
The cutoff is calculated, as I understand it, to be whichever is largest, long-range force cutoff, or 2x the radius of the largest particle plus the skin depth? When I run simulations with similar characteristics to the one provided below I get the output:
not quite. the neighborlist cutoff is determined by the largest cutoff reported by the pair style(s) plus the neighbor list skin.
the fact, that you cannot use the multi communication style is because you are doing a granular system simulation which needs a special neighborlist construction and for which there is no “multi” neighbor list variant programmed. but a multi cutoff variant is not needed, since the neighbor list style for granular media does not use the global cutoff when constructing the list of neighbors, but will consider the two radii of each pair of atoms (plus skin) for constructing the list of neighbors (so you better not choose the neighbor list skin too small.
the only place where the global cutoff is going to be used is for the communication cutoff.
axel.
Ah, thank you for the explanation of the neighboring process. I guess then, my question becomes, is it possible to reduce the ghost atom cutoff to below the 2x largest particle plus skin depth number? As I understand it, the comm_modify command will only allow you to increase the ghost cutoff.
Ah, thank you for the explanation of the neighboring process. I guess then, my question becomes, is it possible to reduce the ghost atom cutoff to below the 2x largest particle plus skin depth number? As I understand it, the comm_modify command will only allow you to increase the ghost cutoff.
not without modifying the source code.
the real culprit, however, is the maximum cutoff needed as estimated in the pair style.
your case is a very special situation. if you had two large particles, the current communication setting would be required for correct force computation.
the current algorithm assumes that you have more than one atom of the largest size.
however, there could be a small chance to have it handled correctly, if you assign a different atom type to the large particle than to all other particles.
atom types are generally ignored, but in the case of estimating the needed cutoff, it may make a difference.
axel.