Sunday 22 April 2007

Molecular modelling tutorial

I was asked to write something about producing the graphics I am putting on this blog. This post will be about the molecular modelling part. The next one about the graphics.

First you have to pick a nice molecule that caught your interest. I'll stick with ligands. N,N'-Diadamantyl-imidazol-2-ylidene looks kind of cool and I am wondering if modeling carbenes works well. To start out you can do the same things that I am doing with any smaller molecule.

For drawing the basic structure, a typical chemical structure drawing program works best. I like ChemSketch. Other people like ISIS/draw. In there you draw the structure. In my case it looks something like this (the c-hexane rings are 3D-rotated):
As a next step you should add the hydrogens in ChemSketch (this works better than adding them in ArgusLab). For quick qualitative results ChemSketch has an optimisation function.

For molecular modelling with more options you can use ArgusLab. To do that you have to export the structure as a molecule file (.mol, .pdb, or .xyz) and open this file in ArgusLab.

Next you should go for "clean hybridisation" (Ctrl+B). The ring becomes aromatic. After that you set up a geometry optimisation (benzene ring button). For a first optimisation molecular mechanics work best. UFF (universal force field) works, AMBER doesn't for some reason. This is the molecule after UFF.

As a next step ArgusLab offers semiempirical QM for structure optimisation. AM1 (Austin model 1) and PM3 (parametrised method 3) are supposed to be the best ones. In this case I used AM1. If you use a molecule this size optimisation takes quite a while. But that is ok because ArgusLab perfectly works in the background.

At first glance the optimisation doesn't make much of a difference. In the following two pictures it can be seen that bondlengths did change a little bit. An interesting fact is that the second structure which was optimised with QM works better with the resonance structures. The bond in the imidazole that is a single bond in every resonance structure really is longer than the other ones.


TINKER is another molecular modelling package that is based on force fields. But I haven't really figured out what to do with it.

If you want quantitative results, you have to go for ab initio quantum mechanics. For that there is no Shareware program and you need much more computing power. The typical program for this is Gaussian. It is definitely a great program but it has a bad reputation for treating competition in a non-academic way (by academic I mean putting science before profit). If you ever feel like modeling a solid body there is my school's WIEN2k. WIEN2k wouldn't ever prohibit anyone from purchasing a license. They even distribute their source code and everyone using WIEN2k can work on improving it. That's what I mean by academic.



Apparently GAMESS is an ab initio program that you can obtain for free. I had been at their homepage before but I was kind of scared then because they only talked about UNIX and self-compiling. But there are also precompiled windows versions for boring half-assed-computers people like myself. I just asked for my registration, I am pretty excited.

5 comments:

Anonymous said...

There's also GAMESS which is free, source and binaries. Its like Gaussian.

Felix said...

thanks. GAMESS seems cool. now that I read through the whole thing I found out that they have a precompiled windows version. I just ordered that

Ψ*Ψ said...

Hmmm. Free, you say?

Felix said...

the only price you have to pay is that you have to read through that whole homepage to find out where you can ask for a downloading permission.

I just received my permission. I will download it and try it out as soon as I have some time

ping said...

PC-Gamess (or Firefly) is also free and optimized for intel cpu based Windows PC.