On Mon, Jan 06, 2025, Adrian Roitberg via AMBER wrote:
>
>I was chasing this problem, was not able to find the root or fix it,
>and then other things came up that distracted me.
My brief summary:
1. I recommend setting ntpr=1, so you can see precisely what is happening.
OK..that creates a big mdout file, but you can delete it when the analysis
is done.
2. In my trials, the temperature spikes are indeed at a *single* time step.
The background temperature and kinteic energy profile (aside from the single
step spike) is very smooth. To my eye, there is no indication that anything
is perturbed in the following steps (i.e. even in the very next step). So
it behaves as if the "bad" temperature value is just something printed, but
never actually used in the simulation.
3. I was never able to reproduce Charo's observation that there were
correlated bond energy changes. This doesn't mean they don't exist, just
that my limited explorations saw the temperature spikes, but not the other
features.
4. I don't think this behavior is limited to (say) multi-heme protein
simulations. It shows up in a small, stable protein like bbl as well.
See the two attached plots, from a bbl simulation. The first shows a single
point temperature spike; the second shows the bond energy over the same time
range. The bond energy is fluctuating (as it should), but it is very
smooth, with no indication that anything happened at the central point where
the temperature spike appears.
If there are temperature spikes that don't look like these examples, that
might help narrow down the problem.
....dac
_______________________________________________
AMBER mailing list
AMBER.ambermd.org
http://lists.ambermd.org/mailman/listinfo/amber
Received on Mon Jan 06 2025 - 12:30:02 PST
