Doing DFT is fun: it’s fast, easy, black-box, and you never know if your numbers are “predictive”. Sometime your favorite functional works, sometime it fails miserably (and you hide the results). Deal with it!
In this post, I’m not going to present any calculations. I want to propose my new TV show!
Now DFT is in its mid-life crisis, as famous theoretical chemists are fighting if “Density functional theory is straying from the path toward the exact functional” or not. This paper (on Science!) of Medvedev (I don’t know him-he is not famous), together with John P. Perdew (everybody knows him-he is extremely famous) tested a bunch of DFT functionals and showed that DFT is straying from the path toward the exact functional. Many people (including me) criticized it because (a) the authors used a test set of only closed-shell atoms, and (b) it “bashed” modern functionals, mainly the Minnesota family. The paper received a lot of attention and citations. Some people like it because they can use it to support their paper employing old functionals such as PBE, PBE0, or B3LYP. Some people hate it does not support them. Kepp (I met this guy in a conference, he is quite cocky IMHO) has the fastest response by publishing a paper (also on Science!), also using a bunch of functionals studying a bunch of atoms. He then criticized Medvedev et al. and showed that “DFT is not straying from the path toward the exact functional”. The second response comes from Korth (on Angewandte), also criticizing the Science paper. In a response to Keep, Medvelev et al. presented their “ideal” way of building a functional and stressed that “unconstrained fitting of flexible functional forms can make a density functional more interpolative but less widely predictive”. So far, there is no response from Truhlar at all. Please, do something!
Now, my proposal: let’s create a Jersey Shore TV series for DFT, in which scientists work together and “expose” each others and the audiences are poor PhD students, who are lost now. We have something similar such as Faraday discussions, but people are too polite there.
A Jersey Shore TV series for DFT?
Finally, to poor PhD students, learn from Keep, start doing benchmark (now!) the electron density for a bunch of molecules, using a bunch of functionals. You might have a chance publishing your results on Nature,
Sciene, Angewandte, or JACS. For the ones applying DFT to study real chemical systems, just listen to your supervisor, you want to graduate, right? After that, you can do whatever you want.