[gmx-developers] Changed dummy generation in TYR

[Erik Lindahl]

Hi,

> 
> Hmm. It already was a pretty dirty construction (or, weird at least),
> but this sounds even worse... Though I agree, it probably isn't so bad
> an approximation, compared to the rest. The rotational inertia could
> be tuned by adjusting the mass distribution over the oxygen and virtual
> mass particles (it doesn't need to be half-half).
> 

No - I've tried to put 25% on the virtal mass, and that works perfectly 
too. It doesn't matter that the mass is really small since no forces 
apply directly to it.
Using 25% on the virtual mass will make the rotational inertia around
the CG-OH vector correct, but the center-of-mass of the hydrogen will be
too close to the OH atom.

You're probably more experience on this than I; the only reason I ended 
up choosing to reproduce the c-o-m instead of inertia was because that's
what you did for all other constructs.

By the way - since I rewrote the dummy generation to use bond lengths
and angles from a database file I also calculate the massses to 
reproduce the center-of-mass exactly for all the rings. I don't think it 
will improve things, but it won't hurt.

> A nicer solution would be to allow a constraint with a dummy atom, but
> I think that would make an additional constraints step necessary. That
> is a bit too much for such a small problem...
> 

That won't work since the dummy forces are interpolated back onto atoms
before we do the update & constraining. And if we use a hard angle it 
will probably oscillate.

If you really want the hydrogen to be a real atom the simplest solution 
is probably to write your own constraining algorithm and apply it to
the TYR HH only before interpolating dummy forces back. Shouldn't be too 
hard since there are only two atoms involved :-)

Cheers,

Erik