In the previous post, I used a bunch of LDA functionals to study the hydrogen bond in the water dimer. The idea is quite simple but unique and “novel”: combine different exchange and correlation functionals and take an average result.
In this post, I update the result with nearly 4000 GGA functionals.
- A lot of functionals predict quite good binding energy, within 1-2 kcal/mol. Many even give me an error of less than 0.2 kcal/mol.
- Some functionals give horribly bad results, either too strong binding, or no binding at all.
- The average number: 5.13±4.66 kcal/mol, it’s a perfect match with the experimental value 5.02 kcal/mol. This is the highest accuracy I have ever since I start doing computational chemistry. MN15, can you do this?
- Using this approach, we don’t need extra correction for dispersion correction such as DFT-D3
- What should I call this “functional”? My colleagues propose me “people functional”, I then call it PEEFUN.