Dear Amber Developers,
Following up on the previous discussion (
http://archive.ambermd.org/202506/0076.html), I conducted additional tests comparing the densities produced by the Berendsen and Monte Carlo barostats in both QM/MM and MM simulations.
In the QM/MM runs, I treated a NO₃⁻ ion as the QM region and 489 methanol molecules as the MM region, using DFTB3-D3 as the QM method via DFTB+. In the MM runs, both the NO₃⁻ and methanol molecules were treated classically. Each system was simulated for 2 ns, and the average densities were recorded as follows:
Run1-- qmmm Berendsen barostat: 0.7668
Run0-- mm Berendsen barostat: 0.8109
Run3-- qmmm Monte Carlo barostat: 0.8117
Run2-- mm Monte Carlo barostat: 0.8121
These results suggest that the Berendsen barostat tends to underestimate the density in QM/MM simulations. This finding is consistent with my previous observations and reports by others (
http://archive.ambermd.org/201212/0011.html). I hope these additional results help further support the issue.
I am currently performing thermodynamic integration (TI) in which the initial state is QM/MM and the final state is MM. Since this workflow is only supported in Sander (and not pmemd), and Sander does not support the Monte Carlo barostat in combination with TI, I would appreciate any suggestions or workarounds you might recommend. At the moment, it seems that running in the NVT ensemble is my only option.
Best regards,
Zihao Bai
_______________________________________________
AMBER mailing list
AMBER.ambermd.org
http://lists.ambermd.org/mailman/listinfo/amber
Received on Mon Jul 07 2025 - 12:00:15 PDT