Re: [AMBER] ig=-1 for multi production runs

From: Jason Swails <>
Date: Mon, 21 Dec 2015 08:09:40 -0500

On Mon, Dec 21, 2015 at 7:31 AM, Fabian gmail <>

> Dear Amber experts,
> Maybe I have a trivial question, but I would like to know if I do run 10
> ns in say 5 x 2 ns runs, I guess I need to use ig = -1 only in the first
> one or in the equilibration stage only to get different samplings for let’s
> say 3 different such full runs (min + equil + prod) correct?
> I mean I don’t need to use ig=-1 in each of the 5 production runs which
> each of them will use the previous .rst file as the starting velocities and
> positions, so ig is not necessary, as I want to continue with the
> conditions saved in the last .rst of the previous prod.rst file, is that
> correct?

​Sometimes, but usually no, this is wrong. The best ways to maintain
constant temperature are either Langevin dynamics or the Andersen
thermostat, both of which employ a stochastic model to maintain the
temperature. Therefore, if you run with either ntt=2 or ntt=3, not setting
ig=-1 will introduce severe artifacts into your simulations. The same
thing goes with any method using random numbers (REMD, barostat=2, ntt=2,
ntt=3, icnstph=1, icnstph=2 are the ones that come to mind).

Setting ig=-1 is very often a good thing, and never hurts (unless of course
you *need* short-time-scale reproducibility for some testing), so I would
highly advise against omitting it just because you don't think it will do


Jason M. Swails
Rutgers University
Postdoctoral Researcher
AMBER mailing list
Received on Mon Dec 21 2015 - 05:30:04 PST
Custom Search