[lammps-users] instability of respta for npt applied to mixed phase systems

Dear fellow Lammpers,
I have had some frustrating moments with the beloved respa integrator when doing npt simulations of a mixed phase system. Basically the box keeps growing until it has reached cosmic proportions … which is of course surprising and a little unphysical. In contrast when I use the humble verlet integrator, everything is stable, my system converges to the expected equilibrium density. As respa is very beautiful and very fast, this is very unfortunate. By the way for the system we are looking at, NVT/NVE with respa is fine, and reproduces the essentially the same results as verlet applied for the same ensemble. Any ideas of what might be wrong ?
The respa problem does not appear if use respa to simulate water at stp
(as you would expect).

The system consists of charged nafion polymers (which are pretty stiff) , corresponding hydronium molecules (with opposite charges to the polymers) and water. The system is overall charge neutral.

Best Regards,


Are you also using SHAKE? There are some known problems with the pressure
calculation when using SHAKE + rRESPA. Though I haven't seen anything
as severe as what you mention, more like it's off by some amount.


I should add that the only way I think your problem can
occur, is if the pressure you are seeing output with
thermo info is different when you run your system with
vs without rRESPA. This assumes you are using rRESPA
correctly with good timesteps, etc. But it sounds like
you are, due to the NVT/rRESPA system working well.

Computing pressures with rRESPA is somewhat tricky
due to the virial contributions coming from different
pieces of the code at different times. SHAKE compounds
that b/c of how/when the constraints are applied.

So can you verify that you are getting radically different
pressures near the beginning of your run with rRESPA
vs Verlet? This would tend to drive the system to a different