[gmx-users] How do Charge Groups work?

[Berk Hess]

>From: "X.Periole" <X.Periole at rug.nl>
>Reply-To: Discussion list for GROMACS users <gmx-users at gromacs.org>
>To: Discussion list for GROMACS users <gmx-users at gromacs.org>
>Subject: Re: [gmx-users] How do Charge Groups work?
>Date: Wed, 28 Jun 2006 20:29:17 +0200
>
>>I had a system where I have no charges in it, but I do have Lennard-Jones 
>>non-bonded interactions. I've noticed that I get different values for the 
>>Lennard-jones interaction energy differs depending how I split the system 
>>up into charge groups (as the charge group concept is kinda abitrary in a 
>>system with no charges), this can not be right surely?
>>
>>As I understood it the charge groups were used to help compile the 
>>neighbour lists, but what I want is all the atoms within my cut-off radius 
>>to be in my pair-list (for that atom). Is there an option to ignore the 
>>charge groups? I find I get bigger Lennard-Jones interactions if I put my 
>>whole system into a charge group that if I split it up, but I need to 
>>split it up for my larger systems. Is there a way to increase the maximum 
>>number of charge groups? And is this a good idea?
>
>You should check the manual for the option allowing a atom
>base cutoff, I am not sure it actually exist.
>A crude way to fix it would be to define each atom as a
>seprate charge group.

This is not a crude way to fix it, this is the only and proper way to fix 
it.

But you will still have a cut-off and your simulations will become slower
due to more expensive neighbor searching.
It is usually better to use shifted LJ interactions (see the manual),
than you can also use larger "charge" groups again to speed up the
nieghborsearching, provides that rlist is larger than rvdw to account
for the charge group size and diffusion between neighor list updates.

Berk.