[gmx-developers] [gmx-users] Wrong Nose-Hoover integrator

[Berk Hess]

Hi,

I have implemented the correct Nose-Hoover leap-frog scheme as written in:
http://www.nrc.nl/binnenland/artikel/1111643830212.html

But I have not been able to reproduce the incorrect 3.2.1 thermostat work
as reported by Michel.
For a box of 216 waters with delta_t=0.002 and 0.9 nm RF I get correct work
at least up to tau_t=0.1 ps.
Michel, how many waters, what cut-off scheme and what time step did you use?

The correct leap-frog algorithm (or probably any reversible integration 
scheme)
poses another problem.
With dt=2 fs SPC water with a full step tempature of 300 K (determined 
from v(t)),
has a half step temperature of 297.8 K. With dt=4 fs the difference will 
be much larger.
The full-step temperature is the more accurate expression, but can not 
be used
in any simple integration scheme. This means that the temperature for water
simulations with reversible Nose-Hoover coupling will be off by at least 
0.7%.
I am afraid this problem can not be solved.

Berk.