From: Kenno Vanommeslaeghe (kvanomme_at_rx.umaryland.edu)
Date: Sun Sep 28 2014 - 14:00:43 CDT
On 09/28/2014 02:31 PM, Axel Kohlmeyer wrote:
> On Sun, Sep 28, 2014 at 2:06 PM, Kenno Vanommeslaeghe
> <kvanomme_at_rx.umaryland.edu> wrote:
>> No. Large harmonic restraints (i.e. large enough to mostly prevent a bond
>> from stretching) would cause *exactly* the same integration time step issue
>> as large bond force constants; to the integrator, there's no difference. We
>> just had this whole discussion less than a month ago;
>> The NAMD developers didn't react that time, and neither did they on an
>> earlier closely related thread, so I'm henceforward going to assume it is
>> not possible to define rigidBonds between heavy atoms, meaning you'll have
>> to use a different simulation software. If so, dear NAMD developers, I
>> myself have absolutely nothing to gain from it, but there does seem to exist
>> a certain demand for rigidBonds between heavy atoms...
> FYI, for a linear molecule like CO2, this is not enough. you also need
> to constrain the angle, and that is a very tricky issue, since at 180
> degrees, the usual SHAKE algorithm leads to degenerate solutions.
I feared as much, but was entertaining the wishful thought that some shake
implementations might converge with one O-O and two C-O distance constraints.
> there are two straightforward alternatives: 1) use a rigid body
> integrator (also needs a short(er) time step) 2) recast the 3 atomic
> linear molecule into to a diatomic for the purpose of time integration
> using two dummy atoms. then you can use SHAKE (and RATTLE) and a
> fairly large time step. solution 1) is for example implemented in the
> LAMMPS, solution 2) in GROMACS.
FWIW, according to the CHARMM documentation, it supports directly applying
SHAKE on arbitrary bonds, which would presumably solve the rigid
chloroform and "constrain all bonds in a protein" problems. However, I
have little experience with how well this works in practice, especially
when given the above 3 constraints in a CO2 molecule.
This archive was generated by hypermail 2.1.6 : Wed Dec 31 2014 - 23:22:53 CST