[gmx-users] fixed dihedrals

[Christoph Freudenberger]

Anton Feenstra wrote:
> That will *not* keep your dihedral fixed! Within the limits
> of a constraint, there is still considerable freedom for the
> dihedral angle to fluctuate (it's easy to calculate if you
> remember your goneometry lessons). For the Shake constraints
> algorithm, this construction will be hard to converge, and
> for Lincs it will probably not even be possible to use it
> (too many connected constraints).
> 
> So - the only way to do it *is* adding an improper, and *yes*,
> setting the force constant high will give high frequencies in
> the motion in this angle, and can result in integration errors
> and ultimately a simulation crash.
> 
I agree with anything you added.
But is the residual freedom left by the constraints really larger
than that given by the improper?

Of course one has to consider the topology of the total
molecule. IMHO it is a good option to add another constraint
to fix some diherdal, if you use already angle constraints on the
whole molecule. Of course you should only use as few constraints
as possible to keep things easier for SHAKE. LINCS will always
crash sooner or later if coupled angle constraints are used...
in my experience ;-)
If you keep angles and or bonds flexible, improper dihedrals
are the better choice, no doubt about that.

regards
-- 
Christoph Freudenberger
Abt. Organische Chemie I AK Siehl - Uni Ulm -Tel: ++49-731-502-2785