On Mon, Feb 11, 2019, Abdullah Bin Faheem wrote:
>
>I would like to inquire about some behaviour observed while running
>NPT simulations on electrolyte systems by both using single long
>trajectoriesor dividing the long trajectory into numerous smaller runs
>(for e.g. one 1.5 ns run or three 500 ps runs). It seems that at each
>restart the MSD curve shows a "bump", where the MSD curve drops in slope
>at the end. Reducing the collision frequency (gamma_ln) from 5 ps to
>1 ps (for Langevin dynamics at ntt=3), seems to give more similar MSD
>curves however the drop in slope exists. This drop only exists for the
>aforementioned smaller runs. I have attached both MSD's for reference and
>the input file is as follows:
>
>NPT ensemble simulation with temperature scaling
> &cntrl
> imin = 0, irest = 1, ntx = 5, ntpr = 500, ntave = 1000,
> ntwr = 500, ntwx = 100, ntwe = 50, ntf = 2, ntc = 2, ntb = 2,
> cut = 12.0, nsnb = 10, nstlim = 500000, dt = 0.001,
> tempi = 298.0, temp0 = 298.0, ntt = 3, tautp = 1.0,
> ntp = 1, pres0 = 1.0, taup = 1.0, ioutfm=1, gamma_ln=5,
> &end
This has the flavor problems that arise in Langevin dynamics when the same
random number sequence is re-used for multiple runs. What version of
Amber are you using? The current default is (I think) ig=-1, which forces
a new random number seed for each segment. But earlier versions had a
default with a fixed seed. Check your outputs to see what seed is being
used, and make sure it is different for each run.
This may not be the problem, but it's simple to check. If you are
re-using the same seed, add "ig=-1" to your input files.
...dac
_______________________________________________
AMBER mailing list
AMBER.ambermd.org
http://lists.ambermd.org/mailman/listinfo/amber
Received on Mon Feb 11 2019 - 05:30:05 PST
