9
$\begingroup$

I'm thinking about a quick-and-dirty approach to molecular dynamics, mostly for teaching purposes.

At a given temperature, statistical mechanics tells us molecules have translational and rotational and vibrational energy as given by the equipartition theorem. At a basic level, all three motions derive from movement of the individual atoms along Cartesian axes.

So it seems like a very simple quick-and-dirty approximate molecular mechanics could be implemented by "shaking" the individual atoms with mass-weighted random displacements. You'd need to implement some sort of force field to make sure the added displacements would fall into the correct potential energy distribution for a given molecule.

Quick searching doesn't seem to find this as an approach, likely because it won't obey a particular canonical ensemble.

I'm curious... Has anyone implemented such a method for qualitative visualization? What are some of the pitfalls compared to "proper" molecular dynamics? (e.g., presumably ensuring the proper velocities?)

$\endgroup$
  • 3
    $\begingroup$ It sounds suspiciously Brownian. $\endgroup$ – Abel Friedman Jan 1 '15 at 22:29
  • $\begingroup$ Well, you'd use some force field to enforce the bond potentials, angle bending, etc. This would distribute the displacements into normal modes. $\endgroup$ – Geoff Hutchison Jan 1 '15 at 22:31
  • $\begingroup$ Will the students implement such a model themselves or is it a visualisation you provide for them (with background of course)? $\endgroup$ – John H. K. Jan 3 '15 at 16:14
8
$\begingroup$

The random displacements in your suggested method are uncorrelated from step to step. What you are suggesting is essentially heavily damped molecular dynamics at the Brownian limit. The limitations are obvious, you won't sample correlated, large-scale motions efficiently. (That might be a student project right there: turn up the Langevin parameter, see when you lose rotations of phenyl rings, and ask yourself what this means.)

$\endgroup$
  • $\begingroup$ Right, that's a great point. I had a sense that this approach would lead to inconsistency even at the qualitative level. $\endgroup$ – Geoff Hutchison Jan 3 '15 at 17:17
  • $\begingroup$ Your point about tuning the Langevin is interesting from a teaching perspective. Do you know any open source MD that would make it easy to tune on the fly? $\endgroup$ – Geoff Hutchison Jan 3 '15 at 17:19
  • $\begingroup$ @Geoff Hutchison: I'm mostly familiar with NAMD, which is at least "free-as-in-beer". You can set the Langevin coupling coefficient with "langevinDamping", but according to the manual, just once per config file. Several config files and some small amount of unix scripting will be required for that experiment. $\endgroup$ – Abel Friedman Jan 3 '15 at 19:39
4
$\begingroup$

I use Molecular Workbench (http://mw.concord.org/modeler/) when I want to employ MD for teaching. Free of charge and easy to use. No need to reinvent the wheel.

$\endgroup$
  • 2
    $\begingroup$ Hi Jan! The problem I have with MW these days is the Java aspect. It's harder and harder to run Java on the desktop and it's not at all possible on mobile. $\endgroup$ – Geoff Hutchison Jan 3 '15 at 17:15

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.