[gmx-users] CHARMM force field implementation in Gromacs : conversion of periodic into Ryckaert-Bellemans parameters.

[Mark Abraham]

> Hello,
>
> I have noticed that both in the Yuguang Mu's and the Mark Abraham's work,
> the periodic parameters of dihedral angles have been converted into
> Ryckaert-Bellemans ones. I have tried to find more info about this in the
> CHARMM and Gromacs documentations but I have not found much. Why exactly
> this conversion should be done since the periodic potential is implemented
> in both force fields? My problem is that several dihedral angles cannot be
> easily converted in RB parameters since their multiplicities is equal to 6
> and the RB potential implemetation is limited to 5 constants.

To quote my own code comment,

"# We need some elaborate functionality to convert the CHARMM dihedral type
# of k * (1 + cos(n * xi - delta ) ) functions summed over n into something
# GROMACS can implement. While the above functional form exists in
# GROMACS, you can't have more than one function of this type, and
# CHARMM has a number of dihedral interactions that require more than
# one such function. However for delta = 0 or pi and n <= 5, then the above
# cosine function can be expanded in powers of cos xi, and the coefficients
# of the expansion can be summed in this conversion and presented to
# GROMACS as a ready-made Ryckaert-Bellemans dihedral. In practice, this
# works because CHARMM only uses such delta and n values for atom type
# combinations that need multiple functions of the above form. Warnings
# are issued when delta is some other value, and the algorithm dies if
# n is > 6. In order to simplify GROMACS logfile output so that it only
# has to report one sort of dihedral term for most simulations, all
# dihedral terms with n <= 5 are expressed as R-B, even when not necessary.
# Dihedrals with n=6 are left in periodic form, since it is not possible
# to convert these to R-B form when the summation is limited to the
# fifth power of cos xi."

So if you have a single dihedral over a set of atoms that has n>=6 then
you can leave it in periodic form and the only cost is that you have to
remember that the output will likely have both periodic and R-B dihedrals.
If you have one such a dihedral in combination with others n<6 then you
can use a combination of periodic and R-B. If you have multiple dihedrals
with n>=6 you will need to hack the source code, except in some trivial
cases, perhaps.

Mark