[gmx-users] fluorine parameters in GROMACS

[Erik Lindahl]

Makoto Yoneya wrote:
> Dear GROMACS users:
> 
> I'm trying to run MD of fluorine containing molecule with GROMACS-3.0.5.
> But, I'm confusing the difference of fluorine LJ (c6,c12) parameters between
> GROMACS,
> GROMOS and related literature.
> 
> The standard GROMACS force fields (ffgmxnb.itp) lists as
> [ atomtypes ]
>       ;name        mass      charge   ptype            c6           c12
>           F    18.99840       0.000       A   0.78301E-02   0.25209E-04,
> and GROMOS (ffG43b1nb.itp) lists as
>           F     0.000      0.000     A  0.0011778624  7.6073284e-07.
> These c6 and c12 are largely different each other.
> Some literatures from GROMACS group lists,
>           F                                A   1.177862E-03   1.000000E-06
> in Fioroni,M. et. al J.Phys.Chem,B'00 104, and
>           F                                A   1.177862E-03   1.168000E-06
> in Fioroni,M. et. al J.Phys.Chem,B'01 105.
> These are close to that of GROMOS as the author says.
> And fluorine parameters in CVFF force fields (e.g. Koike & Yoneya
> J.C.P.,105,'96) are,
>           F                                     0.9847354E-03
> 0.8419906E-06
> and again rather similar to GROMOS ones.
> GROMACS manual says the F parameters are from van Buuren & Berendsen,
> Biopolymers 33, '93.
> In this literature, epsilon and sigma for fluorine was listed as 66.61K and
> 0.305nm,
> and these gives c6 and c12 as,
>           F                                A   1.78334E-03   1.43560E-06.
> These values are rather similar to GROMOS parameters not to GROMACS.
> I'm wondering where are the GROMACS LJ parameters, i.e.,
>           F    18.99840       0.000       A   0.78301E-02   0.25209E-04,
> comes from and are these largely different valued parameters valid for
> standard MD?

Hi Makoto,

I hope David can give you the details on the parameters; The variation 
you describe is actually quite common - even if the c6 and c12 
parameters can be very different, the combined Lennard-Jones potentials 
will probably be quite close to each other in the range where it's 
significant, i.e. 0.1-1.0 nm. I've experienced this myself when looking 
at sodium/potassium ions. In general, all parameters will probably 
provide reasonable pair correlation functions, but they are optimized 
for different properties, like relative free energies of hydration or 
the relative variation of sigma/epsilon in the periodic system.

If you want to use your own set of parameters, you can just copy & 
rename all the files for a forcefield (e.g. ffgmx*), edit the F 
parameters, and just change the #include statement in the topology to 
use your own force field instead.

Cheers,

Erik