Updating diffusion constant midway through the simulation run

User discussions
1

GROMACS version: 2024
GROMACS modification: No

I have been running an AWH simulation for 1 us before realising that the diffusion constant is perhaps too low, due to the lack of convergence. Is it ‘allowed’ to update the diffusion constant midway through the simulation run? The simulation was run with four walkers and the histograms equilibrated at t = 7230 ps.

2

No. But there is also no point. The diffusion constant, together with the initial error, set the initial update size. You can only set the initial update size at the start of a simulation. The diffusion does not need to match the real diffusion. If it is off you can recompute the actual initial error you used using the initial update size formula.

There is usually only an issue when you set the diffusion constant (or initial error) much too low, then convergence can be slow. To remedy that you will have to start a new simulation with larger initial update size.

3

Thanks a lot, makes sense! I will try benchmarking the initial error - I can see that the initial coverings are not very extensive.