Hello, everyone and thanks for helping me.
My question is:
I have a metal complex that has several possible conformations. I need to calculate what conformation is the most plausible in different solvents and maybe the distribution of these conformations.
Will it be ok, if I first optimize the geometry of the first 100 conformations by MM methods, than 50 conformations by PM7 and then 20 conformations with PBE or PBE0 then I'll calculate their final energies? Will I be able to input these energies into Boltzmann distributions? Is COSMO model good enough for this?
End of question
Clarification why do I need it.
Our team is studying clathrochelates (metal complexes with the metal incapsulated in the ligand cage) and we've found the following effect of optical activity induction. When a chiral moiety is presented in the clathrochelate structure, a pike in CD spectra of different chromophore becomes visible (that chromophore is metalopolyhedra) that has different shape in different solvents (sometimes inverted by x axis)
I think it would be nice if I could predict these spectra by calculating the distribution of conformations in given solvent, then calculate their CD.
Dear Serhii,
your question is the very focus of some of my Labmates. For the conformational aspect of TM complexes, I can highly recommend the expertise of Markus Bursch . He recently gave a nice talk on his results, which are published in this paper: https://pubs.rsc.org/en/content/articlelanding/2021/cp/d0cp04696e
The bottom line from this work is that in the semi-empirical world, GFN2-xTB works best for TM complexes (much better than PMx methods), and GFN-FF seems to be the best force-field. However, neither of those is accurate enough for reliable conformational ranking. For this purpose, you need to do some decent DFT calculations. On the cheap end, they suggest B97-3c (if I remember correctly). Even single-points on xTB structures are a major improvement.
The best method for conformational ranking concerning the accuracy/cost-ratio is presumably r²SCAN-3c, which we have recently published here (see Fig. 7 in particular).Preprint r²SCAN-3c: A "Swiss army knife" composite electronic-structure method
The only problem is that is not widely available (yet) but implementations in ORCA and Turbomole are under way.
For conformational searches in general, I recommend the CREST software, which is developed by Philipp Pracht: Article Automated Exploration of the low-energy Chemical Space with ...
The other aspect you mention (CD spectra) is/was the topic of the PhD of another one of my colleagues, namely Jakob Seibert (was because its Dr. Jakob Seibert now:). I think all three of them would be happy if you show some interest in their work, but you will find many useful results in their recent articles.
Cheers!
Jan
Dear Jan,
Thank you very much, this is truly helpful. I was advised about CREST and I'm currently exploring it, but GFN2-xTB and r²SCAN-3c - I didn't know about them. I'm certainly gonna go through all of the articles you've pinned.
I have a different approach to this problem. Let me start by saying theoretical conformational distributions do not correspond to experimental because of solvent effects. One way of achieving your objective is to determine a reasonable number of possible conformers (usually the lowest-energy five conformers) by theory. Use these conformers to simulate their NMR to generate coupling constants. These theoretical coupling constants can be obtained by DFT NMR calculation or using the Altona equations. Perform a linear regression of the experimental coupling constants (which is an average of all conformers present in solution) against those generated with each of the conformers. This should produce the relative population of each conformer, check if the sum of the conformers corresponds to unity. This technique was used in this paper: J. Org. Chem.2013, 78, 5, 2005–2011. Cheers!
- Badges
- Science topic