[lammps-users] Antw: Re: bug in NVT integrator?

Steve, Axel

Thank you for the good hints. I have found a nice paper (you probably know it) of Stephan C. Harvey (S. C. Harvey, R. Tan, T. E. Cheatham, J. Comp. Chem. 19/7, 726 (1997)) discussing the phenomena of the flying ice-cube, and showing possibilities to get rid of the errors in the energy equipartition, done by the thermostatting.
Maybe you should post the citation in the tutorial with the hint that for long simualtion times one has to take special care for the particle trajectories.

I started now some simulations removing periodically the system’s total angular and linear momentum. I tried also the recommanded method from the paper, applying a stochastic thermostat, using fix langevin. But then the total energy is strongly fluctuating, much stronger than with NVT. Is it ok to use fix langevin without removing the totale momentum, because in the tutorial there was something written about removing the bias (the linear momentum) before the temperature is adjusted.

If it is necessary, how can I compute the total momentum and then feed via fix_modify into fix_langevin?

Best regards and nice weekend


Steve Plimpton [email protected] 5/14/2009 4:40 PM >>>
compute c_watertemp water temp
compute c_phetemp phenylalanine temp

Leave the leading c_ off these 2 lines.


Read the doc pages about thermostats and biasing, section 4.16.