[gmx-users] Re: The sum of the two largest charge group radii is larger than rlist - rvdw (because rlist < rvdw)

[Joakim Jämbeck]

Dear Chris,

We have used the following settings in our recent work (http://pubs.rsc.org/en/content/articlelanding/2013/CP/C3CP44472D):

coulombtype=pme
rcoulomb=1.0
rlist=1.0
vdw-type=cut-off
rvdw=1.4
dispcorr=enerpres

And this works fine without any notes etc from grompp, and you do not have to use the switching function. Should probably change the md.mdp file on the website...thanks for bringing this to our attention!

A colleague of mine have used Slipids with a shorter rvdw and the bilayer properties were still very stable. When I use rcoulomb=rvdw=1.0 I see no major difference either except for a relatively big boost in performance. But please double check that your simulations are able to reproduce experimental observables if changing rvdw to around 1.0 nm., but it should be fine.

Best regards,
Joakim



Date: Tue, 26 Feb 2013 19:56:12 +0000
From: Christopher Neale <chris.neale at mail.utoronto.ca>
Subject: [gmx-users] The sum of the two largest charge group radii is
	larger than rlist - rvdw (because rlist < rvdw)
To: "gmx-users at gromacs.org" <gmx-users at gromacs.org>


> Dear users:
> 
> I am experimenting with the "Stockholm" lipid parameters (Slipids). I downloaded the recommended .mdp file from the developers of this force field (http://people.su.se/~jjm/Stockholm_Lipids/Downloads_files/md.mdp ) and was surprised to see the treatment of non-bonded interactions:
> 
> coulombtype = pme 
> rcoulomb = 1.0
> rlist = 1.0
> rlistlong = 1.6
> rvdw = 1.5
> rvdw-switch = 1.4
> vdw-type = switch
> 
> it struck me as strange to have rlist < rvdw (I didn't even know that this was allowed, but I tried it and it is)
> 
> (Note that this is from their website, but their original paper would have used rcoulomb = 1.4 and rlist = 1.4).
> 
> When I run grompp version 4.5.5 with these parameters, I get the following note:
> 
> NOTE 2 [file empty.mdp]:
>  The sum of the two largest charge group radii (0.000000) is larger than
>  rlist (1.000000) - rvdw (1.500000)
> 
> This system has only one Na and one Cl ion (chosen for simplicity during mdp testing). By moving these ions apart and doing a 0-step mdrun, I was able to verify that the LJ interaction energy moves from the LJ(SR) to the LJ(LR) component for d>=1 nm and that this energy is non-zero until d>=1.5 nm.
> 
> I am asking if anybody sees a problem because (a) grompp threw a note, and (b) even if this works alright for standard simulations, I just want to be sure that if I get into more exotic parts of the gromacs code then I am not likely to encounter situations in which this cutoff scheme introduces problems (silent or otherwise).
> 
> Thank you,
> Chris.