Talk:Thermostats

From wiki.gromacs.org

There are certainly cases for using component-wise temperature coupling. In particular with RF (or, shudder, straight cut-off) even using group-based truncation you are supposed to get differential heating and cooling of solvent and solute. http://amber.scripps.edu/Questions/ewald.html also suggests Berendsen pressure coupling, large timesteps and large SHAKE tolerances are bad in this regard. On the other hand, that site also worries that with PME with the same T-coupling for solute and solvent (which have different numbers of DOF) you reduce KE fluctuations more in the solute and slow down its barrier-crossing.

Thus, I think we should be encouraging T-coupling for RF on groups like Solute-Other or Solute-Lipid-Other, and obviously counter-ions get combined with waters, etc. There's a case for the same under PME under some conditions, so we should mention that with riders. Mabraham 03:09, 23 May 2007 (CEST)