Hello,
Recently I recalculate the thermochemistry of a reaction with CCSD(T)/CBS and the results are more lower in energy than how I assumed. The original information was with B3LYP/6-311++G(2d,2p).
My question is: Can I consider my data better? CBS basis set, it isn't very difficult (for small systems) but it didn't seem to be very applied in general., have some consideration or problem that I should know about this extrapolation?
Thank you for your time
As far as I know, CCSD(T) in large basis set is more relible for energy. It would be helpful to comparare these results with predictions of any composite scheme (CBS-QB3, Gn etc) or with MP2-DFT methods (B2PLYP).
Hello,
Thank for your answers. My scheme was, optimization structure with B3PW91/Aug-CC-pVTZ and rest for example ( Hº-E(B3PW91)). Then I calculate energy with CCSD(T) with aug-cc-(D,T,Q)ZV, extrapolate and add to the result of other calculation. giving H(CBS)= [Hº-E(B3PW91)]+[E(cbs)].
The problem is that it is the fist time I use this method and I am a little lost.
I would advise against the use of DZ basis set in the CBS extrapolation scheme, as this possibly introduces errors. In your case, extrapolate only from T and Q basis. Also, you should extrapolate the HF and correlation energy separately. For HF energy use the 3-point extrapolation (D basis included) via the exponential formula, for correlation energy use the X^(-3) formula, which I assume you are using now. And it is not completely clear from what you have written here, but you should be using extrapolation schemes on the total energies of the individual systems, not directly on reaction energies/enthalpies, as they might converge to the CBS limit at very different rates, just in case.
To really explore this more you need to consider a reaction that is well-characterised - one with reliable experimental results. Using B3LYP/6-311++G(2d,2p) as your benchmark is not a good idea.
Just one additional pint. You are not mentioning what kind of systems you are studying. If the wave function of these systems can be represented well with a single determinant then methods based on single determinant can give a good answer. In this case, there is no doubt that CCSD(T) is the benchmark and then you have to see if any of the xc functional works well. In my opinion there are other xc functional that give reliable answer with respect to the B3LYP.
Thank you very much for your answers. I will continue with some test, but now I have an idea the way to go.
Perhaps you could also consider using the explicitly correlated approach (F12-CCSD(T)), which at the cc-pVTZ level should be fairly close to the CBS limit.
This could shed some additional light on the energetics.
Nonetheless, in my opinion including the zero pint energy is mandatory in any quantitative calculations.
Good luck!
Marcin
Joaquim,
Don't forget to take a look on D1 or T1 criteria given by CCSD(T) calculations. There are some systems that are not well represented by a single configuration. In these cases you need to do some tests using CASSCF.
- Badges
- Science topic