[gmx-users] Anisotropic United Atom model and dummy atoms.

[Frederic Leroy]

Dear GROMACS users,

my question deals with the way of modelling an anisotropic united atom 
(AUA) using GROMACS dummy atom option.
The AUA model for a methyl group is made of a mass centre without 
interactions having the mass of CH3 and a masseless interaction centre 
displaced from the mass center at a short distance. The interactions are 
purely LJ.
I think that the 3fad dummy atom is the solution using an angle of 180 
and a displacement of 0.021nm in my case.
I added dummy atoms in the .itp file, adding them or not to the 
constraints, bending, torsional interaction lists. I introduced the LJ 
parameters in the force-field for the dummies and put the two LJ 
parameters of the non-interacting mass centres to 0.0.
I do not have any error message during the "grompp_d" stage or during a 
constant volume run. But the result that I obtained for the total energy 
is inconsistent. The LJ energy is repulsive, as huge as if the dummy 
atoms had LJ parameters greater than they are while I am far from a 
dense state of the system. Thus the converged density strongly 
underestimates the best one the model is expected to yield.
Does anybody have an idea where I made some mistake(s)?

Thanks,
Frédéric, Tarragona Univ.