[gmx-users] v-rescale reproducible restart

Sergio Perez sperezconesa at gmail.com
Thu Mar 7 12:44:30 CET 2019

Dear gmx comunity,

I have done the following:
-Ran a 100ps simulation and then starting  with the same conditions a 30ps
simulation and then restarted upto 100ps.
-If I use -reprod and Nosé hoover thermostat in both trajectories the final
structures match.
-If I use v-rescale it doesn't.

I know since v-rescale is stocastic, random numbers should be properly
saved in the cpt for the chain of random numbers to be reproducible when
doing a restart. I am not sure if this is not happening because it is
impossible to do or if it is a bug. If it is not a bug, a note or something
in this regard in the output would be nice. I am using gromacs.2018.4 with

In addition, I think the name of the thermostat "v-rescale" is not very
good (although I am not sure if it's too late to be changed). It gives the
impression that velocities are just being rescaled (which would be
completely wrong). Furthermore, if we credit with the author name  the
Parrinello-Rahman, Nose-Hoover, Berendsen methods. Why not

Thank you very much!

Sergio Pérez-Conesa

More information about the gromacs.org_gmx-users mailing list