On Sun, Dec 30, 2012, Ismail, Mohd F. wrote:
> This is my input file. Can you comment if there is any bad parameter?
>
> production run
> &cntrl
> imin=0, irest=1, ntx=5,
> nstlim=40000000, dt=0.0005, ntc=1,
> ntpr=2000, ntwx=2000, ntwr=-1000000,
> ntwv=2000, ntwe=2000,
> ntt=0, tempi=300.0, temp0=300.0, ntp=0, ntb=1,
> cut=10, iwrap=1,
> &end
Looks OK to me; you might experiment to see what turning on SHAKE does
(set tol=0.000001 or so). Also, set dsum_tol to 10**-6 (in the &ewald
namelist); regardless of shake, that should help with energy
conservation.
>
> The system contains 32500 atoms.
> > The energy drift is linear, which makes me wonder why. If the system is
> > not fully equilibrated, I expect it to be random. The change in energy is
> > like 20kcal/mol for 20ns. Is this bad?
I don't think this is bad. Note that 20 kcal/mol is a tiny fraction of a
degree for a system with 32500 * 3 degrees of freedom. So your drift
seems within the range of what people usually live with. To put it in
terms Andreas used (see snippet below) you are seeing 1.5 x 10**-9
kT/ns/dof, which is smaller than that often seen, maybe because your
cutoff is 10 Ang rather than 8 (the default).
....dac
>From Andreas Goetz:
The energy drift will depend on details of the simulation (energy
conservation is actually not a trivial topic). Using AMBER, for explicit
solvent calculations using PME, I have typically seen fairly linear
energy drifts +between 1.E-07 (0.5fs time step, no SHAKE) and 1.E-05
(2.0fs time step, with SHAKE) in units of kT/ns/dof at 300K (dof = degree
of freedom). This translates to something between 0.5 kcal/mol and 25
kcal/mol for 20ns for a system with +100k atoms.
_______________________________________________
AMBER mailing list
AMBER.ambermd.org
http://lists.ambermd.org/mailman/listinfo/amber
Received on Sun Dec 30 2012 - 06:30:02 PST
