NVE simulation with energy conservation but not temperature equilibration

Hi all

Sorry for the Long message but I tried to be as clear as possible

I am trying to simulate the heat exchange between a cold system
(ice, lactoferrin, silica nanocube and silica sheet) at 50 K, immersed
in a room temperature water bath (10 times more atoms in water
that in the system to simulate a good bath). I am using charmm FF
and tip3 water. The time step is 1fs, and the systems in water were
energy minimized using SD and CG.

I successfully thermalize each system and its bath using fix NVT for every
group of atoms (I define two groups, one the system and one the water).
First, I thermalize everything to 50 K during 4ps and then, I keep the system
at 50 K and the water at 300 K using two fix for 20ps.

After that, I run an NVE simulation, and compute the temperature for the
system and the water bath. For the ice and lactorerrin (roughly 10K atoms)
the final temperature goes up, and the water bath a little down, both
reach the same temperature (obviously with some larger fluctuations in the
temperature of the system), after a some of picoseconds, around 100 ps.
Unfortunately, for the silica nanocube (10K atoms) and silica sheet (40K atoms),
once the temperatures converge and settle down, there is a large temperature
gap between the temperature of the system and the temperature of
the bath (more than 10 K), which made no much sense for me.

This is really strange for me since the NVE simulations have an energy fluctuation-
variation below the 0.05%. I am using a fine FFT order 6 and a
large cut-off radius (inner 14 A and outer 17 A) for VdW and Coulomb

***I will appreciate any suggestion to get the right temperature and improve energy
conservation.

I add part of my input script

Thanks

John Michael

Hi all

Sorry for the Long message but I tried to be as clear as possible

I am trying to simulate the heat exchange between a cold system
(ice, lactoferrin, silica nanocube and silica sheet) at 50 K, immersed
in a room temperature water bath (10 times more atoms in water
that in the system to simulate a good bath). I am using charmm FF
and tip3 water. The time step is 1fs, and the systems in water were
energy minimized using SD and CG.

I successfully thermalize each system and its bath using fix NVT for every
group of atoms (I define two groups, one the system and one the water).
First, I thermalize everything to 50 K during 4ps and then, I keep the
system
at 50 K and the water at 300 K using two fix for 20ps.

20-24 ps is an extremely short time for equilibration. also, why use a
uniform distribution and why remove the rotational momentum and not
the translational (the latter is often the bigger problem).
also, a dissipative thermostat (langevin or temp/csvr) is usually a
better choice to reach initial equipartitioning.

After that, I run an NVE simulation, and compute the temperature for the
system and the water bath. For the ice and lactorerrin (roughly 10K atoms)
the final temperature goes up, and the water bath a little down, both
reach the same temperature (obviously with some larger fluctuations in the
temperature of the system), after a some of picoseconds, around 100 ps.
Unfortunately, for the silica nanocube (10K atoms) and silica sheet (40K
atoms),
once the temperatures converge and settle down, there is a large temperature
gap between the temperature of the system and the temperature of
the bath (more than 10 K), which made no much sense for me.

This is really strange for me since the NVE simulations have an energy
fluctuation-
variation below the 0.05%. I am using a fine FFT order 6 and a

sorry, your PPPM energy convergence is rather on the sloppy side.
something like 1.e-6 would be more conservative.

large cut-off radius (inner 14 A and outer 17 A) for VdW and Coulomb

with PPPM and narrow convergence, the choice of coulomb cutoff should
have no impact on accuracy only performance, as it determines how the
calculation is split between real space and kspace. it is thus more a
parameter that should be adjusted for optimal performance.

finally, there is no need to use multi neighbor lists, your particle
sizes and cutoffs are quite comparable.

***I will appreciate any suggestion to get the right temperature and improve
energy
conservation.

have you talked to somebody with more experience in MD simulations
than you, specifically with water?

axel.