MEAM + run 0 with multiple elements issue

LAMMPS LAMMPS Mailing List Mirror
1

Hi all,

I was performing quick checks of crystal structures and noticed a particular issue. I construct a two element system, then perform a run 0 to compute the system pressure and energy.

If I use a MEAM potential (from the LAMMPS Potentials folder), I get 1.#QNAN errors for some of the atomic forces (and system pressures). If I deform the system slightly using a change box with remap, no errors. Returning the system to its initial state causes the errors to reappear.

I also tested single element MEAM, and a two element eam/fs and received no errors.

Below is a script that highlights this issue. I’m using the Windows 64 bit 2014-09-09 version of LAMMPS.

Thanks,

Lucas

boundary p p p

units metal

atom_style atomic

variable alat0 equal 3.12

lattice custom ${alat0} a1 1.0 0.0 0.0 a2 0.0 1.0 0.0 a3 0.0 0.0 1.0 &

basis 0.0 0.0 0.0 basis 0.5 0.5 0.5

region box block 0 3 0 3 0 3

create_box 2 box

create_atoms 1 box basis 2 2

#Test 1: 2 element system with MEAM gives #QNAN forces and pressures

pair_style meam

pair_coeff * * library.meam AlS FeS AlSiMgCuFe.meam AlS FeS

mass 1 26.981539

mass 2 58.933195

#Test 2: 1 element system with MEAM works fine

#pair_style meam

#pair_coeff * * library.meam AlS AlSiMgCuFe.meam AlS AlS

#mass * 26.981539

#Test 3: 2 element system with eam/fs works fine

#pair_style eam/fs

#pair_coeff * * AlFe_mm.eam.fs Al Fe

#mass 1 26.981539

#mass 2 58.933195

variable lx0 equal $(lx)

variable deltax equal 0.00001

thermo_style custom step lx ly lz pxx pyy pzz pyz pxz pxy pe

thermo_modify format float %.13e

dump dump1 all custom 100000 atom.1 id type x y z fx fy fz

run 0

undump dump1

variable xmax equal (1+{deltax})*{lx0}

change_box all x final 0 ${xmax} remap units box

dump dump2 all custom 100000 atom.2 id type x y z fx fy fz

run 0

undump dump2

change_box all x final 0 ${lx0} remap units box

dump dump3 all custom 100000 atom.3 id type x y z fx fy fz

run 0

undump dump3

2

Hi all,

I was performing quick checks of crystal structures and noticed a particular
issue. I construct a two element system, then perform a run 0 to compute the
system pressure and energy.

If I use a MEAM potential (from the LAMMPS Potentials folder), I get 1.#QNAN
errors for some of the atomic forces (and system pressures). If I deform the
system slightly using a change box with remap, no errors. Returning the
system to its initial state causes the errors to reappear.

I also tested single element MEAM, and a two element eam/fs and received no
errors.

Below is a script that highlights this issue. I’m using the Windows 64 bit
2014-09-09 version of LAMMPS.

i cannot reproduce it with the current version of LAMMPS.
did you compile LAMMPS by yourself or are you using a pre-compiled executable?

axel.

3

Axel,

I used the lammps windows executable from http://rpm.lammps.org/windows.html website. Interestingly, the version that I was using is no longer listed there. I can reproduce the error with the 2014-09-26 version, but not with the newest version 2015-03-27. Whatever the problem, it seems to have gone away sometime in the last six months.

Lucas

4

Axel,

I used the lammps windows executable from http://rpm.lammps.org/windows.html website.

ok.

Interestingly, the version that I was using is no longer listed there.

that has no special meaning. the installer packages (and all other
precompiled LAMMPS versions) are quite larger and the machine hosting
them is old and small. so there is a script that regularly removes
some of the older precompiled packages.

I can reproduce the error with the 2014-09-26 version, but not with the newest version 2015-03-27. Whatever the problem, it seems to have gone away sometime in the last six months.

most likely is that it was a compiler issue or something unrelated to
MEAM, since there have not been any changes of note in the MEAM code
since sep 1st 2014.

axel.