It has been described a bad modelling parameters for some ions in the AMBER FF. In simulations with high concentration of these ions, they tend to aggregate instead of remaining totally disolved (as expected at the concentrations worked).
This paper describes this fenomena and gives new parameters that improve the behaviour of these atoms: Cheatham's JPCB 2008
Although this paper is from 2008, Amber9 Package release didn't incorporate the new parameters. Thus we should take into account this when setting up simulations at high salt concentration.
SOLUTION
First idea would be to prepare an frcmod file for our ions with the data presented by Cheatham and load it in Leap when preparing the system.But the thing is that Amber10 is already out and they prepared those frcmod files for us that can also be used with Amber9. Download the tar file at:
http://ambermd.org/dbase.html
There you find all the parameters for Amber10 and between them: frcmod.ionsjc_tip3p
Instructions on how to use it here:
http://archive.ambermd.org/201001/0368.html
One must select the parameter's file depending on the water model used, then load it with leap. A recomendation: as we work commonly with TIP3PBOX, it's a good idea to include these lines in the leaprc file:
loadoff ions08.lib
loadamberparams frcmod.ionsjc_tip3p
Managing the paths corretly, of course..
;)