I am trying to equilibrate the system using NVT and NVE ensembles, following previous literature. The literature suggests running the equilibration for 200,000 to 400,000 steps. My question is: Is it necessary to run the simulation for so long if I am achieving constant properties around 10,000 to 20,000 steps? will it impact the end results?
Hi @Mrinal,
Without more information on your system and goals, it is impossible to tell. Reaching equilibrium depends on both the physics of your system on the one hand and simulation parameters, like the timestep value, on the other, and all that is also constrained by what you want to achieve with your simulation.
If the literature you reference mentions using 200,000 steps, which might not be that big in most cases, they may have good reasons to do so. Otherwise, you are the most qualified person that can test and justify if less is already enough.
1 Like
The problem with reaching equilibrium is that you can only say for certain when a system is not in equilibrium. A system may look as if it has reached equilibrium and may still be in a meta-stable state and then, suddenly during your production run, drop into a lower energy state that may or may not be the ultimate equilibrium. Due to the restricted time and length scales for MD simulation, every choice about when a system is in equilibrium is always a compromise between computational effort and accuracy and reliability of the results. What is therefore a reasonable choice depends on many factors and most of those are specific to the system and protocol used.
2 Likes