Total energy not conserved in NVE simulation with TIP4P cutoff model

LAMMPS LAMMPS Mailing List Mirror
1

Dear LAMMPS users,

I have a question concerned with the TIP4P cutoff model. I wonder why the total energy in NVE simulation with TIP4P cutoff model is not conserved. When I did NVE simulation with TIP4P cutoff model (12A for L-J cutoff and 8.5A for Coulombic cutoff), the total energy was not conserved. In stead, the total energy kept increasing. The potential energy kept constant, but the kinetic energy (temperature) kept increasing. And temperature even reached 10,000K.

However, when I did the same NVE simulation with TIP4P-Ew model, the total energy is conserved. And the temperature keeps constant. I wonder what makes the TIP4P cutoff model different. Thanks.

Attached are the input script and data file for the TIP4P cutoff model. Thanks.

Regards,

Kira Banks

in.water (710 Bytes)

water2000.data (612 KB)

2

Any potential with a cutoff (e.g. LJ or Coulombic), will
not conserve energy exactly, even over short time
scales, as atom pairs move in/out of the cutoff, unless
you have some shifting or smoothing at the cutoff. The
Coulomb part in TIP4P/cut does not, and I'm guessing
you are not doing it for the LJ part either. As to why the
energy increases over time, that could be too large a timestep,
or too long a run, among other reasons. I am guessing
you need a smaller timestep with TIP4P/cut than with
TIP4P/long.

Steve

3

Dear Steve,

Thanks for your advice. After reducing the timestep and applying a shift for LJ, the results become reasonable: the total energy keeps almost constant and so does the temperature. I have another question. Is it possible to shift the Coulombic interaction in LAMMPS? The pair_modify shift yes command seems to only work with LJ interaction. Thanks.

Bests,

Kira

4

Is it possible to shift the Coulombic interaction in LAMMPS?

Not for TIP4P/cut. That's what TIP4P/long is for.

Steve

5

Dear Steve,

But TIP4P long needs a kspace ppm calculation. I cannot use kspace_style pppm/tip4p because this style does not support shrink-wrapping boundary and I have to use a shrink-wrapping boundary. Is pppm the only method to account for the long range Coulombic force? Thanks a lot for your time and help.

Bests,

Han Hu

6

Dear Steve,

But TIP4P long needs a kspace ppm calculation. I cannot use kspace_style
pppm/tip4p because this style does not support shrink-wrapping boundary and
I have to use a shrink-wrapping boundary. Is pppm the only method to account
for the long range Coulombic force? Thanks a lot for your time and help.

no. lammps also supports msm. but for that you'd have to adapt the
lj/cut/tip4p/long pair style (which is straightforward to do). just
compare other coul/msm to coul/long pair styles.

axel.