Chemical Forums

Chemistry Forums for Students => Physical Chemistry Forum => Topic started by: Avicenna on March 25, 2007, 08:40:10 PM

Title: Molecular Modelling Questions? o_O
Post by: Avicenna on March 25, 2007, 08:40:10 PM
Hello to everyone on my first post. I have got two questions one of which is generally about molecular modelling and one is a specific question about my project.

I am normally a genetics student who also happens to take lots of computer courses (yes I am an ab initio not a wet lab kind of guy). I am interested in a serious computational career (not just using the programs but actually developing them). I also happen to be interested in molecular modelling and am taking courses on this topic. But as I said in future I actually would like to find my-self in a position where I can develop such programs my-self. Molecular modelling courses only teach how to use these programs, not the principles that they work on. So I need your advises on this matter. To get to the theory of molecular modelling and molecular interractions I am planning to get physical chemistry and classical mechanics courses (some of which I already have taken). I am most interested in molecular mechanics as I said but semi-emprical methods also draw my attention, so I am also planning to get some courses on quantum mechanics (by this time I hope I am not dead). Does this seem like a right path to any of you experienced molecular modellers d: And also is there any community or forum who is majorly founded upon molecular modelling?

Secondly as I said I am doing a project on molecular modelling, deatils of which are unnecessary but it is about chlorophylls. As you know chlorophylls contain alternating double bonds in its ring structure (suggesting overlapping pi orbitals). Now as you know molecular mechanics and semi-empirical geometry optimizations usually give very similar structures. But not in case, in molecular mechanics optimization, the rings of the chlorophylls are optimized to bent rings (as much as their pi orbitals would let them). But in semi-empirical method they are optimized into straight ring structures. Now this makes me think; am I doing something wrong or it is because molecular mechanics can not compute overlapping pi orbitals so the ring bends as much as its double bonds would allow it?

Thanks, take care d:
Title: Re: Molecular Modelling Questions? o_O
Post by: Ψ×Ψ on March 25, 2007, 11:41:00 PM
MM does nothing at all with orbitals--it looks at the atoms as balls on springs.  It's super-quick, though, and it's not bad if you're just looking at sterics and the like.  DFT is also pretty quick and can give good results for some things.  Currently a lot of people seem to be using it on aromatics.  Ab initio calculations take for-freaking-ever unless you are looking at something very, very small.  The validity of the results and the time the calculation takes depends on the size of the basis set you use.
I got a pretty decent introduction to computational chemistry in my quantum class, but this seems to be the exception rather than the rule (quantum here is nothing short of pretty awesome).  I'd suggest quantum chemistry rather than quantum physics.
Title: Re: Molecular Modelling Questions? o_O
Post by: Avicenna on March 26, 2007, 03:31:45 AM
I know the basics on which SE and MM works on, the first one one force fields that are derived purely from classical physics and use constants taken from experiments. And SE based on methods which calculated outer orbitals wave functions and take the rest as parameteres derived from experiments again where ab initio is soley computational on all aspects.

But I am curious whether if my speculation about the bent ring on mm calculations is correct or not? As the result to this question may change the path and relatedly the time required for my project.

And what is DFT by the way? Is it a MM force field or a SE method?
Title: Re: Molecular Modelling Questions? o_O
Post by: Ψ×Ψ on March 26, 2007, 08:03:06 PM
I'm not sure about the bent ring...it seems possible.  If it had been ab initio rather than MM, I'd say the bent structure represented a local minimum on the PES.  This happens sometimes.  Starting with something closer to the correct structure will give you better results in that case.  Felix (http://chemical-quantum-images.blogspot.com/) probably knows better than I would.  As for DFT (density functional theory)...I'm under the impression that it's "none of the above." 
Title: Re: Molecular Modelling Questions? o_O
Post by: allanf on March 29, 2007, 03:36:39 PM
DFT is both "none of the above" and a Semi-Empirical method.  Its a general method of quantum chemistry and so there are flavors that are pure ab initio, and ones that are semi-empirical as well.  The main claim to fame of DFT is that instead of finding energy in terms of the wavefunction (orbitals) it does so in terms of the electron density, and there are some mathematical tricks that can make this worthwhile.

There's an easy intro to the theory on arxiv that my quantum chemistry prof pointed me to, if you're interested in the fundamentals:
http://arxiv.org/abs/cond-mat/0211443

Also, I'd say you're right that the bent ring is an artifact of molecular mechanics.  The chlorin ring is largely aromatic, and so should be planar.  That this shows up in qm but not mm just shows how important non-local bonding is (i.e. the aromatic "cloud" we're all used to thinking about on benzene rings and the like).
Title: Re: Molecular Modelling Questions? o_O
Post by: Ψ×Ψ on March 29, 2007, 09:15:21 PM
Allan: Thanks for the DFT heads-up!   :)
Title: Re: Molecular Modelling Questions? o_O
Post by: FeLiXe on March 31, 2007, 01:30:20 PM
I am also just a starting molecular modeller so i don't really know too much. it seems that MM has problems when it comes to conjugation. for example phenol has the OH hydrogen in the same plane as the Cs when you do it semiempirically. the hydrogen points away if you use molecular mechanics
Title: Re: Molecular Modelling Questions? o_O
Post by: allanf on March 31, 2007, 02:07:54 PM
There are ways to make a given force-field better at dealing with conjugated systems.  Some forcefields include an atom-type that is aromatic, and most newer force-fields incorporate parameters for common aromatic structures (e.g. benzene) to help deal with this. 

However, to deal with conjugated substructures that are not already incorporated can be rather involved.  A good way of dealing with this is to to a QM simulation of the molecule or moiety to get the bond orders.  Most programs like GAUSSIAN or GAMESS will calculate bond orders by default.  Then you can plug these into the force-field and it should represent the conjugated system better.