Stability for long NVE run

User discussions
,
1

GROMACS version: 2024-dev-20241213-e782eff84f-dirty-unknown
GROMACS modification: Yes/No

I’ve implemented the Path MD algorithm (MD in trajectory space) in a GROMACS copy based on the above version. It’s branching off from the commit that integrates the multi-simulation feature.

Now, I’m testing the implementation for correctness i.e., proper equilibration in NVT and total energy conservation in NVE, using the system acetone in explicit water (TIP3P) with 50k atoms.

I’ve run energy minimization to a state with energy -11013.0 and maximum force 0.7455. To equilibrate, I’m using the sd integrator running 500,000 steps with 5 step annealing control and final temperature 150. For the NVE run, I’m using the md integrator running 5,000,000 steps. For both runs, I’m not using DD.

Now, the NVE run seems to be unstable, crashing at step 2,615,753 with LINCS warnings. The graph looks like this:

image
image2219×1333 118 KB

I would be very grateful for any tips/wisdoms on how to prepare for and successfully run long NVE simulations such that the energy is conserved.

Note: I’ll gladly provide more (specific) information on request.

Thanks in advance,
Nitin Malapally

2

Dear @nitinmalapally

Is the same behaviour present in a clean GROMACS installation (i.e. without your modification)?

3

@obZehn I’m testing the behaviour of the new feature itself. In theory, it must be similar to standard MD, so I guess I’m asking for tips for preparing long NVE simulations.