[gmx-users] The meaning of rlist and rcoulomb for coulombtype=reaction-field

[Mark Abraham]

>>> Does this 0.25 nm buffer get automagically added to the specified
>>> rcoulomb to determine the neighbor searching radius?
>>
>> No, I checked, and that doesn't happen in src/mdlib/ns.c
>>
>> My guess is that the actual implementation includes all atoms in charge
>> groups that have at least one atom inside the respective rlist and
>> rcoulomb cut-offs, and that accordingly the documentation is out of line,
>> but I don't care enough to read ns.c that closely to find out, so
>> hopefully David/Berk/Erik can clarify here.

> It's center of geometry that is use as the criterion. For large charge 
> groups this can be problem,, and therefore you want to add roughly the 
> size of one charge group to the neighboursearching length.

So to clarify, the user should account for the size of their own charge 
groups in specifying rlist and actually specify rlist = rlist_wanted + 
radius_of_largest_charge_group. This ensures there are no holes inside 
rlist_wanted, but the edge of the sphere can still be ragged outside that.

This would also be a good reason not to use large charge groups!

Mark