GROMACS version: 2023
GROMACS modification: No
I am using PME electrostatics for equilibrating a solvent box with <5000 atoms. Is it necessary to put fourierspacing with a definite value? I see people adding different fourierspacing values for a given FF while many people just don’t add this parameter in mdp file so it’s a bit confusing. Should I explicitly add this in mdp?
The default value of fourier-spacing is 0.12 nm. This is ok with a 0.9 nm cut-off and slightly conservative with a 1.0 nm cut-off. For larger cut-off distances increasing it will improve performance significantly.
Unfortunately we still do not have full automation of the PME parameters in GROMACS based on an error tolerance (this turns out to be somewhat difficult).
Thank you very much @hess . This makes things a bit clearer. Will it be OK if I just go ahead with the default value of 0.12 nm when rcoulomb >1.0 (say GROMOS FF, where rcoulomb = 1.4)? If it is just about performance, I think the small system size in my case (~5000 atoms) wont be an issue. However, if there are any other reasons (accuracy, stability of simulation, etc.) I will consider tuning it for my specific cases. Could you please help me on this.
Thanks,
Raman
The fourier-spacing scales with cut-off for roughly constant accuracy. So you could use something like 0.17. But you can also keep 0.12.
Thanks a lot for the quick reply and clarification.