[gmx-users] Gromos DPPC bilayer: different results for 4.0.7 and 4.6.5

Ollila Samuli samuli.ollila at aalto.fi
Wed Dec 18 21:26:18 CET 2013


>>On Dec 18, 2013, at 8:50 AM, Sabine Reisser <sabine.reisser at kit.edu> wrote:
> >Up to Gromacs 4.0.7, the long rance forces, which are calculated at every nstlist'th step, were added constantly
>>to the short range force in the subsequent steps. Starting from Gromacs 4.5., the long range force is added 
>>nstlist times to the short range force on the nstlist'th step, while there is nothing added to the short range force 
>>on the subsequent steps. This creates some kind of impulse force which we believe is responsible for the 
>>different behaviour of the membrane, although we haven't figured out the exact mechanism yet.

>That's interesting. Now I see that this change is documented in the manual (Section 3.4.7). Would that imply that >this effect does not depend on the model for long-range electrostatics (Reaction Field or PME), but it comes from >using twin range cutoffs?

I also noticed the same section in the manual today. I think that what possibly happens is that with the reaction field the forces between rlist and rcoulumb are not used in the steps between neighbour list update. However, in these steps the reaction field force calculation assumes the dielectric envinronment beoynd rcoulomb, not beoynd rlist even though the forces only up to rlist are used. I am not sure if this would have any practical relevance, though? I am also not sure how this is handled in the pressure calculation?

I think that the reason why Patrick did not get the difference in the PME results between versions might be that he probably used rcoulomb=rlist in those simulations. If I try to set rcoulomb>rlist with PME, I get error:

  With coulombtype = PME, rcoulomb must be equal to rlist
  If you want optimal energy conservation or exact integration use PME-Switch

If it is true what I am saying, then the area per molecules should be the same between versions also using RF and setting rlist=rcoulomb.

Samuli Ollila

>> I expect you to get the 'right' area per lipid if you use nstlist 1 as already suggested by Mirco, because in that 
>>case the multistep algorithms give the same answer. This is also our experience with G54A7.

>I started a run yesterday with nstlist=1, and so far (about 12 ns) the area per lipid indeed stays at the "correct" >value of around 0.63 nm^2.



Gromacs Users mailing list

* Please search the archive at http://www.gromacs.org/Support/Mailing_Lists/GMX-Users_List before posting!

* Can't post? Read http://www.gromacs.org/Support/Mailing_Lists

* For (un)subscribe requests visit
https://maillist.sys.kth.se/mailman/listinfo/gromacs.org_gmx-users or send a mail to gmx-users-request at gromacs.org.

More information about the gromacs.org_gmx-users mailing list