When is it time to mothball Density functional theory? [sorry for the second pun..]
Yesterday I had a nice meeting with Charlie Campbell and members of his group at the University of Washington. It was fascinating to see their lab (as a theorist it is always a reality check!) . Over the past decade they have developed several high resolution microcalorimeters which allow accurate determination of the binding energies of different atoms and molecules to specific surfaces. These results present a significant challenge/benchmark for electronic structure methods (such as density functional theory) which claim to be able to calculate accurately such quantities. The results are also of fundamental importance for understanding mechanisms of heterogeneous catalysis.
I found results for adsorption of benzene and napthalene on Pt particularly interesting. They are summarised in the Figure above. DFT gets a binding energy which is too small by almost a factor of three.
I discovered my friends Jeff Reimers and Noel Hush are co-authors of a paper, Adsorption of Benzene on Copper, Gold, and Silver surfaces, where they do a systematic comparison of DFT with higher level quantum chemisty. [complete-active-space self-consistent field theory with second-order Møller−Plesset perturbation corrections (CASPT2) for the interaction of benzene with a Cu13 cluster model for the Cu(110) surface].
A couple of other interesting things they state:
For all systems, the bonding is found to be purely dispersive in nature with minimal covalent character.....
this cluster [of 13 Cu atoms] is actually too reactive and provides a poor chemical model for the system.
A fundamental question arises as to whether DFT is gets the wrong results just because it is bad at dispersion forces or whether there are other strongly correlation effects at play. I would have thought the latter since both benzene is strongly correlated and Pt is moderately correlated that there may be more to it than dispersion forces.
Below are the measured adsorption energies of different alkanes on three different surfaces. The straight lines suggest that there is also a well defined binding energy per carbon atom.