Hi all,
I posted a few weeks ago about running into negative temperatures in some
QM/MM simulations in a protein. I've isolated the problem to be ibelly with
four-point water models. I made a box of 500 waters and ran 100 fs of MM
NVE, dt=1fs, tempi=50 for OPC, OPC3, TIP4PEw, and TIP3P with increasing
numbers of water molecules frozen with ibelly, from 0 up to 475. The
average temperature over the 100 steps for the three-point models remains
consistently about 200 K no matter the number of waters, but the average
temperature for the four-point models increases very high and then plummets
into the negative before climbing back. It looks like some sort of critical
point is reached at about 325 waters frozen out of 500. The reported
kinetic energies are reasonable and decrease as the number of frozen waters
increases, as anticipated. I think there's some sort of bug in the code
that calculates the temperature of a system containing four-point water
models with frozen atoms. In my protein system, I also found that freezing
the protein but not the four-point waters led to positive temperatures, as
expected. Here's a Google Drive link to the data I used to find this,
including the initial structures, sample input file and sander command, and
the csvs and plots of the temperature and kinetic energy vs. the number of
frozen atoms:
https://drive.google.com/drive/folders/1IOEhZT5ZA1lNqlagcz1EOruiSsKn9Wed?usp=sharing
Best,
Matthew
_______________________________________________
AMBER mailing list
AMBER.ambermd.org
http://lists.ambermd.org/mailman/listinfo/amber
Received on Thu Jul 17 2025 - 12:00:03 PDT
