Once again I’ve run into the same ‘ol discussion: why DFT is better than HF. Of course DFT has some advantages over HF in many cases, the inclusion of electron correlation through a parametrized functional being only one of them. What bothers me are the somewhat bogus statements that DFT defenders use; for instance, it’s not hard to listen to someone say DFT is superior to ab initio methods because the electron density is a tangible (?) entity, whereas a wavefunction has no physical meaning. It seems to me-though I’m not a physicist-that its rather common in physics to spend a lot of time trying to find a mathematical entity from which all the physical information of the system can be extracted, such as a partition function (statistical mechanics), a state function (thermodynamics), a Lagrangian (classical mechanics), etc. So what if a wavefunction has no physical meaning per se? That shouldn’t be the issue to choose DFT over HF.
Let us not get carried away by over used methods like B3LYP, which has proven to be a good hybrid functional but in no way is omnipotent as many seem to believe. Every day, new and improved functionals are being published worldwide, the problem is their inclusion in available codes takes a long time. The availability and ease-of-use of many comp. chem. codes is growing more rapidly than the proper instruction of the underlying methods. To me the choice is either a matter of computational convenience or a consequence of what I’m trying to assess at the moment.