[lammps-users] NVE Energy Drift: Rhodopsin Benchmark

LAMMPS LAMMPS Mailing List Mirror
1

Dear Lammps users,

Within the Lammps benchmark folder there contains an input file, in.rhodo, for a Rhodopsin protein in a solvated lipid bilayer.
If I replace the current NPT Integration with that of NVE I see a significant Energy Drift in the Total Energy.

Reducing the time-step to 0.5 fs ( Verlet ) together with improving the tolerance of the SHAKE algorithm has not resolved the issue.

My questions are

  1. What is the source of this Energy Drift.
  2. How can it be resolved.

Advanced thanks for all suggestions,

Jer

2

Dear Lammps users,

Within the Lammps benchmark folder there contains an input file, in.rhodo,
for a Rhodopsin protein in a solvated lipid bilayer.
If I replace the current NPT Integration with that of NVE I see a
significant Energy Drift in the Total Energy.

Reducing the time-step to 0.5 fs ( Verlet ) together with improving the
tolerance of the SHAKE algorithm has not resolved the issue.

going below 1fs time step is pointless with shake. 2.0fs should work fine.

My questions are

1) What is the source of this Energy Drift.

most likely the energy convergence of the long-range electrostatics

2) How can it be resolved.

try: kspace_style pppm 1e-6

axel.

3

I don't know whether this snapshot of the rhodo system
used for the benchmark is well equilibrated. Or that we
are using params that are ideal for a long timescale
protein model. How big is the energy drift? Maybe
Paul can comment on the system. A 2 fmsec timestep
should be fine with SHAKE, as Axel said.

Steve

4

This snapshot is not well-equilibrated and was never intended to be used for more than a computational speed benchmark.

Paul

5

Very good,

Thanks you for your input.

Jer