I would first use parmed to print the matrix of Lennard Jones parameters and verify that there is a nonzero repulsion between the two atoms.
Are you using sander.MPI, the cpu-version of pmemd, or the cuda-enabled version of pmemd? Is this a multistage transformation (decharge, softcore vdW, recharge) or a single-step transformation that uses LJ and electrostatic softcore functions? The available softcore functions depends on the program you are using. If this is a single-step transformation with softcore electrostatics, then I'd recommend using the SSC(2) "smooth-step" softcore function discussed in the manual; however, that is only available in the cuda-enabled version of pmemd. If you use the cpu version of pmemd, then one must resort to using a less-reliable softcore function that can produce situations like you've described. If you must use the cpu version of pmemd, then you'll likely need to change to a multistage strategy, where you separately adjust the charges and van der Waals.
_______________________________________________
AMBER mailing list
AMBER.ambermd.org
http://lists.ambermd.org/mailman/listinfo/amber
Received on Wed Aug 21 2024 - 13:00:02 PDT
