GROMACS version:
GROMACS modification: Yes/No
Here post your question If changing the gen seed value (default is -1) account for different initial velocity then what is the role of Maxwell distribution and gen temp? How the random number generator i.e. seed (numerical value) will be related to initial velocity?
Thanks in advance.
I think the idea is to sample randomly from a Boltzmann distribution the distribution of velocities for your particles. Generally, when you generate (pseudo-) random numbers, you need a “starting point”, that is, a gen_seed to start generating them. If you change the seed, you change the velocities of the individual particles (but if the temperature is the same then the Boltzmann distribution you sample from is the same). So, in principle, changing the seed gives you different starting velocities for your atoms.
The seed is also important for reproducibility purposes, as setting the same seed will give you the same starting velocities in other simulation.
Thank you for the reply. So while performing triplicate, at the time of equilibration(NVT, NPT) and production (in the .mdp file) gen_seed should be changed or only need to change during NVT.
I would generate the velocities only once, when I do the first dynamical simulation after the energy minimization, as for following runs (NVT, NPT etc) I would read the velocities from previous outputs during the gmx grompp compilation by setting gen-vel = no in the mdp file.
If you want to do replicates, then I would generate new velocities for a new replica when I have dynamics for the first time. If you have a standard pipeline of the form energy min. → NVT → NPT → prod, then after EM you can have three NVT sims with different seed and proceed from there.
Thank you.