High accuracy energy methods, are they worth it?

Question

I'm trying to figure out for myself the necessity of high-accuracy compound methods for energy like W1 or Gn families. I'm learning computational chemistry by myself and sometimes lack systematical approach. As far as I understand they are used for high-end kinetics calculations of small molecules.

So, is there a sense to use the more or less "regular" (not focusing on quantitatively modeling complex mechanisms) mechanistic calculations on large organic molecules (~100 atoms) or something like MP2 with large basis set or high-level DFT functional will be more appropriate?

Answer 1

High accuracy energy methods, are they worth it?

They are absolutely worth it for many types of studies, but not worth it for other types of studies. It depends on precisely what the aim of your project is.

As far as I understand they are used for high-end kinetics calculations of small molecules.

They are also used for spectroscopy, benchmarking and for fundamental science.

In this paper of mine the molecule was able to be made thanks to high-precision predictions. The molecule was made by photoassociation (using a photon to associate two atoms into a molecule), for which the frequency of the photons have to be exactly equal to the energy difference between the bound molecular state and the unbound atomic states, and it wouldn't have been possible without ultra-high-precision predictions of this energy difference.

Our high-precision tag here is starting to accumulate quite the collection of interesting questions, for example:

• How accurate are the most accurate calculations? (here the high accuracy ab initio calculations are sometimes more accurate than the best experiments, and you can see this comment chain if you have questions about precision vs accuracy here).
• High precision helium energy (here the high accuracy calculations are helpful as a benchmark, and in testing/developing novel numerical methods).
• Are there examples of ab initio predictions on small molecules without the "major approximations"? (here we can learn some fundamental science about how the Born-Oppenheimer approximation breaks down for very light nuclei and when the precision is very high).

"on large organic molecules (~100 atoms) or something like MP2 with large basis set or high-level DFT functional will be more appropriate?"

Depending on what you want to do, you might even want high-precision energies for your 100-atom system, but it won't be possible to go much further than MP2 for such a large system. It has been possible to do CCSD(T) on 1000 atoms, but it was a very expensive calculation and was only done in a TZ basis set, which means that the composite methods that you mentioned such as the W family and the Gn family, will be limited in how much they can offer; and the accuracy you gain from going from MP2 to CCSD(T) may not be so valuable when there's still so much basis set error anyways.

If you are working on a 100-atom organic system, I would suggest MP2 or a DFT functional that is decent for that type of system, and to use a TZ basis set. Alternatively you can go up to CCSD or CCSD(T) if accuracy is more important to you, and you're willing to wait a few days on a supercomputer for each energy calculation.

It depends very much on what the aim of your project is.