[gmx-developers] Changed dummy generation in TYR

[Erik Lindahl]

Hi,

When I was playing around with dummy support in OPLS I noticed that the 
dummy generation in Tyrosine wasn't always stable.

The problem was that the HH angle was kept constant by introducing a 
constraint to the CG atom. Since this distance is considerably longer 
than the constraint OH-HH it caused frequent instabilities unless you
used lincs_order=8.

I've worked around it by introducing a virtual mass at twice the 
distance from OH (compared to the OH-HH distance). Half the HH mass
is put there, and the other half added to the OH atom. The HH bond
is then recreated as a dummy2 from the OH and virtual mass coordinates.

This will still increase the HH rotational inertia slightly, but it
shouldn't be a worse approximation than anything else of the dummy stuff.

Cheers,

Erik