I am using Gaussian, DFT (probably B3LYP or WB97XD). As an example, if I have an amino acid; any particular one as this is just an example of my compound, and I want to see how the structure and hbonding is affected by different solutions, what is the best way to set up my initial system? Would I build my organic compound then randomly add a 'handful' of explicit molecules which represent my solution? Or should I take more of an educated guess regarding the orientation of the solution to the compound. Would the geometry optimisation performed by Gaussian take care of this though?
Also, unlike in very large MD simulations where I can model accurate concentration of water/solution/ions-to-compound, how many solution molecules in ab initio terms would be enough? I wouldn't want to find that too few molecules leaves the complex 'half' saturated, but then too many and I end up trying to model an almost bulk water likeness which would slow Gaussian right down.
Thanks
Dear Anthony,
That's a tricky question. If you want to see the effect of the solvent on your molecules, may be PCM or COSMO-RS are a good starting point. Of course they are not as accurate as include explicit molecules of solvent. But, despite MD allows including explicit water molecules, I would not compare it with DFT calculations, since they are very different from the theoretical point of view,
One think you can do easily with DFT is combine PCM (or COSMO-RS) model with some explicit water molecules. You should analyse what are the potential hydrogen donors and acceptors in your aminoacid and then place water molecules accordingly. That will give you a "more accurate" view about the interactions between your molecule and the solvent.
I hope I can help you
Goar
Thanks for the reply. It has certainly given me something to think about, especially the hybrid approach of combining a PCM with explicit solvent molecules.
I would suggest including into the Gaussian's geometry optimization as many solvent molecules as necessary to completely saturate the number of possible H-bonds of your aminoacid.
For example, if you've glycine, include at most 3-4 H-bonding donor solvent molecules.
The remaining of the bulk solvent could be simulated with PCM.
We have recently published on the subject of cavity models within the PCM, and the change in default cavity between G03 and G09. If using PCM, it would be beneficial to consider which interactions are taking place, and specifying in the scrf section the "read" keyword, which allows you to add the PCM input section at the bottom of the file. This edits the PCM parameters and is worth using in G03 to ensure that the UFF radii are used (radii=uff). I attach a link to our paper.
http://dx.doi.org/10.1021/jp505344a
Another keyword that you can use in Gaussian is SCRF=(SMD,solvent=xxx).
There are a lot of different solvents available, but for water just replace the xxx with "water".
SMD actually used IEFPCM as the model by default, but is specifically parameterised to calculate dG of solvation. Choice of model depends on the properties you wish to investigate.
- Badges
- Science topic