Starting with same strutue, EM doesn't converged same structure(or step)

User discussions
1

GROMACS version:2021.5
GROMACS modification: No

Hello,
I tried to simulate protein-ligand interaction, PDB 3BUF, with gromacs tutorial for reference.
With ten trials, I found that even I add counter ions with fixed seed (same initial structure), the converge structures and energies were slightly different each other.
When I check the em log files, intial few steps are exactly same each other, but after some iterations they get differ slightly and finally converge on difference step.

What is main reason of this situation? and is there any way to make every minimization converged to same step(or structure)? I use same mdp file of tutorial’s

2

Any MD or EM process may differ slightly due to issues with rounding, order of operations, domain decomposition, etc. The fact that different EM processes converge slightly differently is irrelevant as you can never guarantee that you are in the global minimum, you just need to ensure you are in some favorable state (at least a local minimum).

3

Thank you for your kind reply!