[ultron55]

I have been trying to minimize a clay substance called montmorillonite.
Montmorillonite is an extended solid consisting of 2 layers between which other molecules can be. Each of these layers is made up of alternating silicate tetrahedrals and (in my case) aluminum octahedrals.

http://www.porepressure.info/images/montmorillonite.png

 I have constructed a small part of one of these layers consisting of an aluminum octahedral bonded to 2 silicate tetrahedrals (one on each side) to minimize it so I could obtain vibrational frequencies. The problem is, every time the program runs, my constructed piece falls apart and deforms, especially in vaccum. When ran in aqueous solution, it stays together slightly better, but not well enough. My question is, how can I minimize montmorillonite so that it stays together, how it is found in nature, so that I may obtain frequencies?

The code for the minimization in vaccum I'm using is:
# b3lyp/6-31g(d) scf=direct opt=ModRedun

Aqueous solution:
# b3lyp/6-31g(d) scf=direct opt=ModRedun SCRF=(Solvent=Water)

I realize that charge and multiplicity could be a problem, so I worked out what they should be. Aluminum has a +3 charge, and by my count there are 3 uncounted electrons, so it should have a neutral charge with a multiplicity of 1. However, it is also possible for it to have a -1 charge and a  multiplicity of 1 as well. I have tried running them with both of these options, but to no avail. While I realize that with the -1 1 option, 3 oxygens/hydroxides should detatch and float around the aluminum atom, the whole molecule just loses it's structure. Sometimes it even spits out hydroxides.
I would really appreciate some suggestions, thanks.

[vex]

Have you considered something like periodic boundary conditions? This would essentially replicate the unit cell you're modeling in the form of a potential energy, approximating the effect that nearby unit cells would have. I've only ever seen them implemented in the context of molecular dynamics simulations, but with math and enough processing power, anything is possible. 

[ultron55]

Hmmm... That sounds like it just might work! Unfortunately, I haven't heard of this before, so I have a question. Would this manipulate the bonds and structure of the unit cell to approximate it? I ask because Montmorillonite has a very rigid structure when it comes to it's layers, so it has to be represented in the program in that way.

[vex]

Well, one would hope not. However, electronic structure calculations on extended solids is a bit of a problematic science and the calculations are sometimes very inaccurate. I wouldn't expect wild deformations in your structure if you implement the method correctly, but take your vibrational frequencies with a grain of salt!

[ultron55]

Ok, that makes sense, at least those frequencies are something to go off of for now. Would you happen to have a sample of the code used for one of these calculations? I have been researching this method and looking around for code samples, but the stuff on the gaussian website is a little confusing, almost not well documented.

[vex]

Unfortunately no.  I've never seen it done in Gaussian before, and wouldn't know where to start. Good luck to you though!

[ultron55]

Ok, do you mind telling me which program you saw periodic boundary conditions? Thanks for all your help, I really appreciate it!   

[vex]

I was working with it in the NAMD suite, but that uses molecular dynamics, which is based more on classical physics and won't give you all the information that you're looking for. It might be a good starting point, though, if you take a look at the theory in the NAMD User's Guide.

Sorry for the late response!