[lammps-users] Respa and fix nvt

LAMMPS LAMMPS Mailing List Mirror
1

Dear All,
I am using lammps-20Dec07 version on an Intel cluster.

I am working on a molecular liquid (fully flexible + LJ + Coulomb). I find the “etot” to be constant of motion (1 part in 10^4) for the following two runs.

  1. NVT with run_style verlet (dt=0.5fs)
  2. NVE with “run_style respa” with the following attributes (dt=1.0fs)
    run_style respa 2 2 bond 1 angle 1 dihedral 1 improper 1 pair 2 kspace 2

However, under NVT conditions with respa turned on (same respa break-up as above, and dt=1.0fs), etot is stable (no drift), but with a much larger RMS deviation (around 3 parts in 100). I provide below the input file and the top portion of the parm file for this run.

I would appreciate your inputs.
thanks,
bala
==============INPUT File==========
units real
dimension 3
atom_style full

read_restart run01.rest
include parm.bmimpf6_256

neighbor 1.0 bin
neigh_modify delay 0 every 20 check no

timestep 1.0
run_style respa 2 2 bond 1 angle 1 dihedral 1 improper 1 pair 2 kspace 2

velocity all create 300.0 23482341 mom yes rot yes dist gaussian

fix 1 all nvt 300.0 300.0 500.0
fix_modify 1 energy yes

thermo_style custom step vol temp etotal pe evdwl ecoul elong ebond eangle edihed eimp f_1
thermo 10
thermo_modify flush yes

dump 1 all custom 20000 dump.bmimpf6_256 tag mol type xu yu zu
restart 5000 rA.rest rB.rest
run 20000
write_restart r.rest

2

Bala,

When running NVT, you won't necessarily observe energy conservation
since the thermostat adds or removes energy from the system to
maintain constant T.

For questions about respa and energy conservation, see the literature,
starting with:

Tuckerman, Berne and Martyna, J Chem Phys, 97, p 1990 (1992).

I've been somewhat disappointed with respa's ability to deliver on its
promise to yield good speedup and conserve energy, so I would not
recommend using respa in most cases. I prefer verlet plus SHAKE with a
2 fs timestep for biomolecular and similar simulations.

Paul

3

Hi Paul,
Thanks for your reply. I am aware. If etot of LAMMPS includes the pe and ke from the thermostat degrees of freedom (in addition to that from the atoms), it should be conserved.

anyway, thanks.

bala