Thanks everyone for your help. Now, when I knew what I was looking for,
I was able to find the relevant info in the manual which I
overlooked/did not understand before.
The 'mystery' of the energy drop at restart appears
to be solved, even though I haven't tried the restart using the rstdip
file yet. As for the energy drift during NVE run, however, I did few
more tests and I am now even more puzzled than before:
(1) The original 1 ns NVE run, 1 fs timestep, ipol=1, default for
polarizable potential:
Etot=-6592.0065 for NSTEP=1, then within few ps Etot goes up to
-6575.1718 and then steadily decreases to -6588.2209 at the end of 1ns
run - drop of some 13 kcal/mol from the highest value, however, still
above the initial value for NSTEP=1.
(2) I tried indmeth=1 (in the &ewald section of the input), everything
else the same:
Etot=-6592.0065 for NSTEP=1 (same as above), but then Etot steadily
decreases and drops by more than 200 kcal/mol to -6881.6926 in just
200 ps !!
(3) Then I tried a NVE run with polarizability OFF (ipol=0), starting
from the same initial configuration as (1) and (2). The initial Etot is
of course different and there is a short (10-20 ps) 'equilibration'
period when the energy jumps a bit (within 3 kcal/mol) as the initial
configuration was obtained with polarizable potential, but apart from
that Etot remains oscilating around a pretty much constant value with a
drift of 0.5 kcal/mol at most over 700 ps.
Now, it is very well possible that I overlooked something painfully
obvious and out of ignorance am doing something terribly wrong.
But I am really worried because I have been routinely using AMBER with
polarizable potential for my simulations of slab systems in last few
years. So far I always did constant temperature simulations, and never
noticed anything alarming. However, the strange behavior of constant
energy simulation combined with polarizable potential brings up a
question of how reliable all my previous simulation are. Can it not be
that the velocity scaling due to the temperature control just masks the
problem with polarizable force field for slab systems?
I would be grateful for additional help/comments/suggestions.
best regards
Martina Roeselova
-----------------------------------------------------------------------
The AMBER Mail Reflector
To post, send mail to amber.scripps.edu
To unsubscribe, send "unsubscribe amber" to majordomo.scripps.edu
Received on Mon Feb 02 2004 - 17:53:00 PST
