Have you ever wondered if it is possible to use many different DFT functionals in your paper, not 5, not 10, or 20? I mean more than 100, or more than 1000? Can we use the results and estimate an average number as well as the standard deviation?
In this post, I show that this idea of the “wisdom of the crowd” is not that bad. I calculate the hydrogen bond energy in water dimer with 175 LDA functionals by combining different exchange and correlation functionals in libxc. Here, I only do single point calculations, using CCSD(T) geometries taken from NIST. I show here the binding energy without ZPE correction, i.e. De. The experimental value is 5.02 kcal/mol (taken here)
The results are surprisingly good, even with LDA! (IMHO)
In general, most of the functionals correctly predict the positive biding energy. Moreover, I can now estimate the average binding energy is 6.34±3.29 kcal/mol, only 1.3 kcal/mol too high.
There are more than 4000 combinations of GGA functionals and the calculations are planned for the next post.
PS. I want to write about the baby v2RDM-CASSCF as promised, but the author of this baby
just ignored my question is trying to solve the problem (it has been nearly 2 weeks). Until he answers me, I will continue with the baby.