My epsil value was converted, for example, from 5kj/mol to 1.19 kcal. that's actually the value I have been using.
It's true. I just checked the 2003 Martini paper
(J. Phys. Chem. B 2004, 108, 750-760)
and the attraction between CG water molecule-beads (which make up most
of the particles your system) is 5 kJ/mole which you correctly
converted to 1.19kCal/mole.
My apologies.
To determine whether it is because of 'flying ice cube effects', I am now using fix langevin+fix nph to replace fix npt.
I tried NVE + Langevin (using a somewhat larger volume which was
pre-equilibrated at NPT at a temperature of 310). This did not help.
After 20000 time-steps, the water remained frozen in in a crystalline
state (which resembled the original crystal structure it was prepared
in).
I've never tried the Martini force field. At this point, I'm curious
to see what the problem was. I have the feeling it's something very
simple. Please let us know when you find it.
To determine whether it is because of 'flying ice cube effects', I am now using fix langevin+fix nph to replace fix npt.
I tried NVE + Langevin (using a somewhat larger volume which was
pre-equilibrated at NPT at a temperature of 310). This did not help.
After 20000 time-steps, the water remained frozen in in a crystalline
state (which resembled the original crystal structure it was prepared
in).
to break up a finite size crystal in MD, you usually have
to start with a temperature far above the freezing point.
considering that MARTINI water *will* freeze at much
higher temperatures than regular water, you simply
cannot expect it to melt a 300K (or 310K).
if you simulate long enough after breaking up the crystal,
you may however observe the opposite (hence the need
for "anti-freeze"), and *that* will likely result in a
"flying ice-cube".
I think Axel was right, the freezing point of CG Martini model is surprisingly high, which can be around 300K.
As I have increased the temperature from 310K to 350K. and changed fix npt to fix langevin+fix nph. the system runs well.
I haven't tested the accuracy of the simulation, but at least the particles are not 'freezing' any more.
I think Axel was right, the freezing point of CG Martini model is surprisingly high, which can be around 300K.
it is not really surprising that the melting point is that high.
in martini, you combine 4 waters into one site. that basically
means, that your water is already "pre-frozen". if you use
3 waters per bead the melting point goes down and so on.
this paper of a former co-worker takes a close look at these issues.
Exploring the utility of coarse-grained water models for computational
studies of interfacial systems.,
Xibing He, W. Shinoda, R. DeVane, and Michael L. Klein, Mol. Phys.,
108, 2007-2020 (2010).