What value of tau-p did you use there with PR?
tau_p was 5 ps
I tried to compute the compressibility along x/y for a CHARMM membrane with a protein in it and I get a value of 2-3e-5 bar^-1. So about twice as small as the value for water. This means that the period of PR turns out around sqrt(2) smaller than tau_p. This might cause the issues you see.
Just a curiosity - how do you calculate this? I guess it is 1/ bulk modulus, but not sure how you would get it for a specific FF. Thank you
I ran two NVT simulations with different box sizes. This is straightforward if you want the bulk compressibility. I have not (yet) done the proper calculation for only x/y. I compressed along x/y, but then also the pressure along z goes up.
Thank you! I was wondering if this was some mathematical calculation, which I should be doing as an estimate to start the simulation. I always use default water, as most of my systems are hydrated clay minerals, with ClayFF being based on the SPC/E FF.
Ah, but then the compressibility might be even much smaller!
But you might also be able to compute the compressibility from the period of the box fluctuations when using Parrinello-Rahman. The correction factor for the compressibility is the square root of the actual period divided by tau_p.
Very good point. I actually donāt know. I have been running various clay systems for years and have had no issues with system stability (till the earlier comment). I have had problems with other materials (non-clay) and it has been also tricky to stabilize something like layered double hydroxide (a very thin layer, very flexible/undulating, and always requires very gentle equilibration to prevent it from disintegrating). Experimentally it is a fragile system also. So, possibly, I should look into changing compressibility when working with those.
If this is true for most of the systems (protein-protein, protein-ligandā¦ etc ), can you please update it in your tutorials. It would be a great help for newcomers.
Thanks
None of my tutorials use Parrinello-Rahman in conjunction with velocity generation.
Yes! but when I am using āParrinello-Rahmanā for protein-protein MD simulation, at NPT equilibration step, in a note Gromacs suggests not to use it.
Should we ignore it?